Title of Invention	A MULTI-USER INTERFERENCE CANCELLATION UNIT ANDA METHOD THEREOF
Abstract	The invention disclose a method and unit for multi-user interference cancellation, which method mostly include: the input baseband signal is de-multiplexed and spreaded spectrum utilizing de-spread unit; then the de-spread signal proceeds to be de-spread with Walsh in different channel, so that the signal of each channel for user is separated; the separated bit stream of each channel multiplies a conjugate signal of the output estimated value from a pilot estimator A to counteract the effect in multipath fading; the multiplied bit stream is input to the decision circuit for judging; the steps for the constructing the threshold of judging as required etc.; utilizing other two pilot estimators B, C as well as the number of subscriber, the type of channel and the type of server providing by the system, the invention can control and process the signal decision and signal restore more accurately, increasing accuracy that interference is cancelled thereby improving performance of the system.

Title of Invention

A MULTI-USER INTERFERENCE CANCELLATION UNIT ANDA METHOD THEREOF

Abstract

The invention disclose a method and unit for multi-user interference cancellation, which method mostly include: the input baseband signal is de-multiplexed and spreaded spectrum utilizing de-spread unit; then the de-spread signal proceeds to be de-spread with Walsh in different channel, so that the signal of each channel for user is separated; the separated bit stream of each channel multiplies a conjugate signal of the output estimated value from a pilot estimator A to counteract the effect in multipath fading; the multiplied bit stream is input to the decision circuit for judging; the steps for the constructing the threshold of judging as required etc.; utilizing other two pilot estimators B, C as well as the number of subscriber, the type of channel and the type of server providing by the system, the invention can control and process the signal decision and signal restore more accurately, increasing accuracy that interference is cancelled thereby improving performance of the system.

Full Text	A Method and Unit for Multi-User Interference Cancellation FIELD OF THE INVENTION The invention relates to multi-user detection in a code division multiple access (CDMA) communication system, more particularly, to a method and unit of multi-user interference cancellation. BACKGROUND OF THE INVENTION CDMA is one of multiple-access modulation technologies widely used in mobile communication. Other multiple-access modulation technologies include Time Division Multiple Access (TDMA) and Frequency Division Multiple Access (FDMA). CDMA technology has advantages of high frequency band utilization and large system capacity over other multiple modulation technologies. In CDMA communication system, signals of each user overlap in time domain and frequency domain, which are distinguished only by their spreading codes. If these spreading codes are completely orthogonal to each other, the signals of each user can be completely recovered by using a correlator and matched filter. In practical system, however, the users are not synchronized and the signals of various users arrive at the receiver with various time delays, therefore it is really hard to find out a kind of spreading code sequences that can make the signals of all the users to be orthogonal within all possible relative time delays. Since spreading code sets are not completely orthogonal, there are interferences among various users or various multi-paths of one user, which is so-called multiple access interference (or multi-user interference). A conventional receiver demodulates the signals by using a correlator (or matched filter), but the interference among users could not be cancelled during a correlating process. For a certain user, the signals from other users are all deemed as noise. Therefore, with the increase of users in the system, the interferences among the users increase gradually. When the interferences are cumulated to a certain degree and exceed the minimum SNR required in system modulation, the system is unable to allow more users to access. Therefore, CDMA system is a system of interference limited. To solve the problem of interference limited in a CDMA system, the influence from multi-user interference must be reduced. In fact, unlike system thermal noise, multiple access interference (MAI) can be estimated and regenerated theoretically. Therefore it is possible to reduce the MAI in signals received by integrating useful information of every user and adopting certain signal processing methods, which is the object to be realized in the multi-user detection technology. Effective multi-user detection technology can increase system capacity and system covering radius, and alleviate the problem of interference limited in a CDMA system. Meanwhile, interference cancellation (multi-user detection) technology can effectively reduce the influence of "near-far effect" on the system performance. Because of the influence of wireless channel fading and the difference of the distances between a base station and mobile stations, the signal power of each mobile station received by the base station is different. A user with strong power signal will impose strong interference on a user with weak power signal, so that the performance of the user with weak power signal is reduced, and even may not operate normally. Using power control technology can make almost equal power among all the mobile stations received by base station, and alleviate the "near-far effect" to a certain degree. However, power control still has some disadvantages, such as, using channel to transmit power control information, control lag, performance related to the rate of mobile users, etc. Furthermore, power control could not solve the problem of system capacity limited by MAI. Power control aims at limiting the interference to an acceptable level. Unlike power control, interference cancellation technology is used to eliminate the interferences among the users to a maximum extent, so as to eliminate the cause of "near-far effect" radically. Therefore, the interference cancellation technology can alleviate effectively the influence of "near-far effect", and relieve the performance requirement for power control in the system. Generally, mobile communication faces a time-varying multi-path fading environment. In this environment, the signals transmitted through various paths will arrive at a base station with various time delays. Since the spreading codes are not completely orthogonal, mutual interferences similar to MAI exist among signals of various paths of the same user. In design of a conventional CDMA receiver, Rake branch is adopted to demodulate respectively signals of several paths with the strongest power for each user, and then maximal ratio combination is implemented, so that the effect of multi-path fading is overcome by using diversity reception technique. During demodulating for each path, however, other multi-paths are still deemed as noise, and information containing thereof could not be used. Thus, in the environment of multi-path propagation, multi-access interference and multi-path influence could not be overcome simultaneously only by simple diversity reception and combination. If the multi-path influence is considered in the algorithm of multi-user detection, it is obvious that the effective information can be adequately used and the system performance will be further improved. A multi-user detection receiver can be classified as linear multi-user receiver and nonlinear multi-user receiver depending on their structure with or without feedback. A linear multi-user detector equates MAI in multi-user communication environment with a transmission response matrix of a channel. The transmission matrix is related with spread sequence of each user and the relative time delay between spread sequences of various users. If an inverse matrix of a channel transmission matrix is obtained, multi-user signals can be output through K (K is the number of users) matched filters, and implemented inverse operation by using the matrix, so that the correlation among users can be cancelled equivalently and the object of canceling MAI is achieved. However, in this method, exact phase information between spread codes should be known, and the inverse matrix of related matrix should be calculated momentarily according to any change of practical users or environment. Hence, the algorithm of the method is complicated, the quantity of calculation is great, and real time realization is not easy to approach. The other kind of important multi-user detector is called nonlinear multi-user detector (also called subtract!ve interference cancellation detector). In its basic principle, MAI information from each user is estimated independently in a receiving end, next partial or entire MAI of the corresponding user is subtracted from total receiving signal so as to obtain the interference cancelled signal of each user, and then a traditional receiver is used for demodulation. Generally, the feedback-based detector is realized by multistage mode, and expected that interference is cancelled to the maximum extent by using multistage feedback so as to obtain better demodulation performance. From the analysis on a linear multi-user detector, it can be seen that the linear multi-user detector uses matrix operation with complex calculation, which does not facilitate hardware realization. From the view of realizing, nonlinear multi-user detector is more effective. Nonlinear multi-user detector can be realized by a serial or parallel structure. A detector with a serial structure generally requires firstly ranking input signals according to power, implements interference cancellation to the user with stronger power orderly, takes the signal of interference cancelled as input, and then implemented the same processes to the users with weaker power. Under the situation that the effect of power control is not obvious or is lagged, the performance of serial process is better than that of parallel process, but it will incur process time delay in proportion to the number of users. When the powers of users are relatively balanced among one another or the request of process time delay is higher in the system, the parallel structure is generally adopted. No matter which structure is adopted, multistage processing is commonly implemented to obtain better effect of interference cancellation. No matter whether a serial or parallel structure is adopted, the performance of a multi-user detector is determined by an interference cancellation unit (ICU). ICU generally includes two parts of signal deciding and signal recovering. The object of signal deciding and signal recovering is precisely reconstructing the data of each user at the receiving end so as to cancel interference from other users during interference cancellation and meanwhile to ensure that additional interference will not be incurred from errors in signal reconstructing during interference cancellation. The precision of signal deciding and signal recovering will directly affect on signal reconstruction capability of ICU, which will determine the integrated performance of the detector. Thus, how to implement exactly signal deciding and signal recovering becomes a critical factor for improving detector's performance. At present, the research on multistage interference cancellation is generally classified as non-decision interference cancellation and decision interference cancellation (also called soft-decision interference cancellation and hard-decision interference cancellation). Non-decision interference cancellation uses directly output of a correlation receiver to generate interference recovery signal, during which it does no need to estimate channel parameters. Its algorithm is relatively simple and it can obtain the gain of interference cancellation in white noise channel. However, if the effect of multi-path fading is not considered in Rayleigh channel, the signal obtained without through Rake process can not overcome the effect of fading. In addition, during interference recovery if the multi-path signal passed through channel is not reconstructed, the recovery signal obtained by such process is very different with real receiving signal, which is likely to induce interference after interference cancellation and directly results in reducing the system performance under the Rayleigh environment. Decision interference cancellation implements decision by using the signal Rake-combined and recovers multi-path signal during reconstruction so as to cancel multi-user interference and multi-path interference effectively. According to patent 6081516 of NEC, "Multiuser Receiving Device for Use in A CDMA System", firstly the hard-decision is executed for the data Rake-combined, the multi-path signal is recovered respectively by using path estimation, and then interference cancellation is implemented, which is applicable to a Rayleigh fading environment. However, because the decision is implemented without corresponding process based on the characteristics of signal received, when the received signal amplitude of a certain user or path is less, the reliability of implementing decision and recovery by using the signal is relatively lower. Under this situation, since the interference recovery signal is inaccurate, the interference is induced during interference cancellation. In "Third Generation Mobile Radio Systems Using Wideband CDMA Technology and Interference Canceller for Its Base Station" (see FUJITSU Sci. Tech. J., 34, 1, pp.50-57 (September 1998)), some technologies are adopted to improve the precision of signal decision and signal recovery. A mixed decision is firstly implemented at signal decision. Particularly, a threshold is set according to a energy of path estimate, if the energy of the signal Rake-combined is bigger than the threshold, +1 or -1 is decided; if the energy of the signal Rake-combined is less than the threshold, a value (less than 1) normalized by the threshold is decided. It is understandable that the reliability of received signal is relatively higher when it is bigger than the threshold, and the decision can be deemed precise; when it is less than the threshold, however, the reliability of received signal is worse, and the decision is likely to be an error. If the decision is wrong, the interference will on the contrary be increased during interference cancellation even though the result of decision is a relatively smaller amplitude value. In this way, the accumulation of errors will reduce rapidly the performance during such multistage process. In "Successive Interference Cancellation for Multiuser Asynchronous DS/CDMA Detectors in Multipath Fading Links" (see IEEE TRANSACTIONS ON COMMUNICATIONS, VOL.46, NO.3, MARCH 1998), the method of threshold decision is also adopted. Unlike the above paper, the decision is not implemented when the energy of the combined signal is less than the threshold. In addition, the threshold is calculated according to the energy variance of the received signals from users to be demodulated, the energy variance of users interfering, and energy variance of noise. But in the practical system, it is not a simple process to obtain the precision value of three energy variances. A more complex calculation is needed to determine the threshold related with three parameters according to changes of an environment. SUMMARY OF THE INVENTION The object of the present invention is to provide a method and unit for multi-user interference cancellation to overcome the disadvantages of inaccurate signal decision and signal recovery existing in the current interference cancellation unit. To achieve the above object, the present invention adopts the following technical solution. A method for multi-user interference cancellation according to the present invention includes the steps as follows: a) implementing an operation of de-complex spread for input base band signal by using a de-spread unit b) implementing Walsh de-spread of channels for the de-spread signal to separate the channels of a user; c) multiplying bit streams of each separated channel by a conjugate signal of an output estimate value of a pilot estimator A to cancel the effect of multi-path fading; d) sending the bit streams multiplied to a decision device for deciding; e) constructing a threshold for deciding by using other two pilot estimators B and C and the number of users, channel types and service types provided by system, and then deciding the input bit streams; and f) multiplying the bit streams decided by an estimate value output by the pilot estimator A to reconstruct the effect of multi-path fading and subsequent reconstruction of the signal. In the above step d, when deciding, the following steps are further included: dl) obtaining parameters such as the number of the users, and channel types or service types from the system; d2) determining the information of channel types or service types; d3) adjusting the coefficient K2 in the equation of Threshold = K1.K2.E_B according to the information of the step d2; d4) determining the number of users in the current system; and d5) adjusting the coefficient Kl in the equation in step d3 according to the number of users. In the step d3, when adjusting the coefficient K2, the higher channel rate, the higher the value of K2 is taken. The value of K2 corresponding with convolution code channel should be less than that corresponding with the channel of Turbo code type. When adjusting the coefficients, K2 is determined first, and then Kl is determined. The value of K2 is approximately linear increase with an increase of channel rate and the change of channel code type. In practical operation, Kl can be set provisionally as about 2 to 3, and then to simulate and test the optimal value of K2 under different channel rates and different code types. In the step e, when deciding, a double thresholds decision can be adopted. Namely, when the energy of a bit to be decided is less than threshold 1 (Tl), the value of "0" is decided; when the energy of the bit to be decided is bigger than the threshold 2 (T2), the value of "0" is decided too; the value of "1" or "-1" is decided for other situations. In the step d3, the adjustment also can be implemented according to the equation of When adjusting, first the output signal of the pilot estimator C is obtained; next the measuring base "Bas" for E_B is calculated according to the equation of whether the E_B is bigger than Bas is judged, if true, it is shown that the bit energy is stronger and the value of K3 should be reduced; then whether K3 has been reduced down to the value less than 1 is judged, if true, K3 = 1; if E_B is less than Bas, it is shown that the bit energy is weaker and the value of K3 should be held or increased; and then whether K3 has exceeded a maximum MAX_K3 is judged, in which MAX_K3 is an integer bigger than 1 . A multi-user interference cancellation unit of the present invention includes a de-spread unit, three pilot estimators A, B and C, a first multiplication device, a conjugate device, a decision device, and a second multiplication device; the de-spread unit de-spreads an input signal, and the useful signal of the user is separated from total signal; on the one hand, the de-spread signal is sent to the three pilot estimators, and on the other hand, the de-spread bit stream is sent to the first multiplication device; the output signal of the pilot estimator A is conjugated by the conjugate device; the first multiplication device multiplies the de-spread bit stream by the signal conjugated through the conjugate device; the decision device uses the output signals of the pilot estimators B and C and system information to decide the bit stream to be decided that is output from the first multiplication device; the second multiplication device multiplies the signal output from the decision device by the signal output from the pilot estimator A; the result of multiplying is an input of subsequent stage of a demodulation module. BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a schematic diagram of a traditional receiver; Fig. 2 is a structure schematic diagram of traditional parallel interference cancellation for multi-user detection; Fig. 3 is a structure schematic diagram of an interference cancellation unit according to the present invention; Fig. 4 is a flow chart of an interference cancellation method according to the present invention; Fig. 5 is a process schematic diagram of an interference cancellation method (changing decision threshold according to different situations) of the present invention; Fig. 6 is a demodulation performance curve of a receiver under different decision thresholds according to the method of the present invention; Fig.7 is a process schematic diagram of decision (how to reduce the error rate of decision) in an interference cancellation method of the present invention; Fig.8 is a process schematic diagram of decision (how to determine the confidence of decision) in an interference cancellation method of the present invention. DESCRIPTION OF THE PREFERRED EMBODIMENT(S) Fig. 1 shows a basic principle of a receiver. A RF module converts a RF signal received into a base band signal, and sends the base band signal to a demodulation module for demodulating. The symbol stream demodulated is sent to a decoder for decoding. Generally, a RF processor includes a mixer and a local oscillator, which can convert the RF signal into an intermediate frequency signal, and then further into the base band signal. The demodulation module includes a Rake receive module for searching K pieces of the strongest multi-path signals; a de-spread module uses hybrid PN spread sequence corresponding to each user to de-spread each path signal of each user; finally, the de-spread symbol stream is multiplied by Walsh function sequence corresponding to each channel, and then the bit stream obtained by the multiplying is sent to the decoder for decoding. The decoder chooses related decoding mode according to different coding mode of signals. For example, the fundamental channel uses convolution coding mode, the corresponding decoding mode is Viterbi decoding; while the supplemental channel uses Turbo coding mode, the corresponding decoding mode is Turbo decoding. Fig. 2 shows a parallel interference cancellation (PIC) structure for a traditional multi-user detection. The structure includes M-stages multi-user detection process (M>1). Each stage includes N pieces of ICUs, in which N is the number of users. The N ICUs adopt the parallel arrangement mode and implement the interference cancellation simultaneously, unlike in serial mode interference cancellation in which the strongest user is chosen to implement interference cancellation firstly. The delay length of the delay units 201-1 to 201-M of each stage equals to the process time delay of the corresponding stage, to ensure that the two input signals of the subtracter 204-n (n is 1 ~ M) are synchronous. It is should be noted that the interference cancellation structures of the first stage and the last stage have little difference from that of other stages. The input signal of the first stage is taken from the output of the matched filter 206, while the input signals of other stages are all taken from the output of the subtracter of the preceding stage, namely the signals that the interference cancellation has been implemented. The ICU 2-M-l to 2-M-N of the last stage only need to implement interference signal reconstruction. After the multi-user detection of all stages is finished, the final output signal An is an interference reconstruction signal by subtracting all other users except user n from total signal (n is 1~N). In theory, if the signals of all users can be reconstructed accurately, the demodulation performance even in a multi-user situation is same as that of one user. Fig. 3 is a structure schematic diagram of ICU in Fig.2, namely an ICU structure adopted by the present invention. The unit includes a de-spread unit, three pilot estimators A, B and C, a first multiplication device, a conjugate device, a decision device, and a second multiplication device, the de-spread unit de-spreads an input signal, and separates the useful signal of the user from total signal, on the one hand, the de-spread signal is sent to three pilot estimators, on the other hand, the de-spread bit stream is sent to the first multiplication device, the output signal of pilot estimator A is conjugated by the conjugate device, the first multiplication device multiplies the de-spread bit stream by the signal conjugated through the conjugate device; the decision device uses the output signals of pilot estimators B and C and system information to decide the bit stream to be decided output from the first multiplication device; the second multiplication device multiplies the signal output from the decision device by the signal output from the pilot estimator A, the result of multiplying is an input of demodulation module of the subsequent stage. In a traditional ICU structure, only one pilot estimator is used. The signal of the pilot estimator is used not only for weighting the de-spread signal and reconstructing the received signal, but also as a reference signal of the decision device. However, the essential of weighting the de-spread signal and reconstructing the received signal is to cancel and reconstruct the multi-path fading effect in the received signal, which requests that the pilot estimator should not have an overlong cumulated length and time delay so that it can comprise enough exact multi-path fading components, but this kind of pilot estimate is not applicable to the reference signal of a decision device. Thus, in the ICU structure provided by the present invention, special pilot estimators B and C are added to provide the reference signal of a decision device, while the pilot estimator A is used specially for weighting the de-spread signal and reconstructing the received signal. Fig. 4 is a flow chart of interference cancellation method according to the present invention. In the method of the present invention, in the first step 401, the input signal is de-spread by using the hybrid PN spread sequence of the corresponding user, the useful signal of the user is separated from the total signal, and the de-spread signal is then sent to the pilot estimators A~C. In the second step 402, the Walsh de-spread is implemented for the PN de-spread signal to separate the signals of each channel of each user. The steps 401 and 402 are implemented within the de-spread unit. The steps 404, 408 and 410 extract the pilot estimates A, B and C according to different pilot estimate mode, respectively. In the step 412, the bit stream of each channel separated is multiplied by the conjugate signal of output signal of the pilot estimator A to cancel the effect of multi-path fading, and the bit stream multiplied is sent to the decision device for deciding. In the step 416, a threshold for deciding is constructed by using the other two pilot estimators B and C and the system information provided by step 414, and then the input bit streams to be decided is decided. In the step 418, the bit stream decided is multiplied by the pilot estimate A to reconstruct the effect of multi-path fading, and then the subsequent reconstruction processes of the signal are implemented, such as Walsh spread, hybrid PN sequence spread, which are not described in details here. In the above method, the output of the pilot estimators B and C and system information are all inputs needed for deciding by the decision device, and the detail function of the device will be described below. The function of the decision device is to decide the input bit stream as a sequence of 1 and -1. The most simply decision method is that if the amplitude of a certain bit is bigger than level 0, it is decided as 1, otherwise as -1. This decision method is very inaccurate if considering the effect of thermal noise and multi-path fading. In order to decide more accurately and avoiding additional interference incurred as much as possible, it is necessary for the decision process to introduce a threshold. The new decision method is that if the amplitude of a certain bit is bigger than the threshold, it is decided as 1; if the amplitude is less than the negative threshold, it is decided as -1; level 0 is decided for other situations. In this method, it is considered that the decision confidence of the bit is lower when the bit amplitude is weaker, it is better to decide it as 0 to avoid possible additional interference incurred, rather than a non-zero value. In addition, because of a wild fluctuation of bit stream energy caused by multi-path fading, using a fixed decision threshold will obviously not yield an exact decision result. Meanwhile, it is difficult to predict a value of the bit energy, so a fixed threshold is really hard to realize. Considering the above factors, the threshold of the present invention can be obtained by an equation below: Threshold = K1.K2.E_B (1) in which, E_B is an energy value of output signal of the pilot estimator B, the value of Kl depends on the number of users, the value of K2 depends on the channel type or service type of a corresponding user, and Kl and K2 are non-negative numbers. For example, if the user uses a supplemental channel of 153.6 kbps, its K2 is bigger than a fundamental channel of 9.6 kbps. Kl and K2 are obtained by emulation and real field test. Fig.6 is a curve diagram of demodulation performance obtained by emulation under different decision thresholds of different user numbers, and the channel used is the fundamental channel of 9.6 kbps. It can be seen that the best decision threshold is different for the different number of users. It means that the change of Kl follows the change number of users. The optimum Kl and K2 could be concluded based on a mass of emulation data and real field test data. Fig.5 is a process schematic diagram of interference cancellation method (changing decision threshold according to different situations) according to the present invention. In this process, in the first step 502, the parameters, such as the number of the users and channel type or service type, are obtained from system, and it is easy to obtain the parameters. In the second step 503, the channel type or service type is determined, for example, it is a fundamental channel or supplemental channel, what is the channel rate, and which coding type is adopted. In the third step 504, the coefficient K2 is adjusted according to the information of the step 503, for example, the higher the channel rate, the higher the value of K2; the value of K2 related to the channel of convolution code type is less than that related to the channel of Turbo code type. In general, the value of K2 tends to approximately linearly increase with an increase of channel rate and the change of code type (changing from convolution code to Turbo code). In a practical operation, Kl can be set as from about 2 to about 3 provisionally, and emulate and test the optimum value of K2 under different channel rate and different code type. Because the optimum value does not much relates to number of users, the number of users for testing can be 10 to 40. After the K2 is determined, in the fourth step 505, the current number of users is determined. In the fifth step 506, Kl is adjusted according to the number of users. For example, referring to Fig.6, when the number of users ranges from 10 to 20, the optimum Kl is 1; when the number of users is 30, the optimum Kl is 2; when the number of users is 40, the optimum Kl is 4 or 5. It is easy to determine the range of Kl by emulating, which should be not too big or too small. Fig.6 is an emulation result only under a special environment. A mass of emulation and real field test should be done to obtain the optimum Kl adapted to various environments. Although the optimum Kl may offset in different communication environment, the offset is not too big. Therefore, a balance point could be chosen among the optimum values to basically satisfy various environments. Such balance point is the optimum value of Kl adapted to the number of current users. In practical realizing, a table of Kl and K2 could be made according to testing data, in which the system can query the related Kl and K2 according to real system information (the parameters such as the number of uses, the channel type, code type). The E_B in the above equation (1) uses an energy value of the pilot B. The traditional ICU has only one pilot estimator A. Because the pilot estimator needs to weight the bit stream before and after decision to cancel and reconstruct the effect of multi-path fading, the time delay and accumulated length of the pilot estimator should not be overlong. However, the variance of the signal output from such pilot estimator is bigger and the energy fluctuation is wilder, which is not suitable to generate the decision threshold. Therefore the pilot estimator B should be added to do the work. The difference between the pilot estimator B and A is that the accumulated length of the pilot estimator B is longer, then variance of the output signal is smaller and energy fluctuation is less. Additionally the envelope of its energy satisfies basically the energy envelope of the bit stream to be decided, so it is suitable to generate the decision threshold. From the emulation test, it can be seen that compared the effect of using only the pilot estimator A with using the combining pilot estimator A and B, the demodulation effect is improved obviously by using the pilot estimator B to generate the decision threshold. When deciding, if the bit energy is weaker, the decision confidence is lower. With the same principle, if the bit energy is too strong over a certain limitation, the decision confidence of the bit is also obviously lower because the energy will be deemed as an abnormal energy resulted from interference. Based on the above discussion, the present invention adopts double-threshold decision to further reduce the additional interference caused by wrong decisions. In the solution of the present invention, when the energy of the bit to be decided is less than the threshold 1 or bigger than the threshold 2, the bit should be decided as 0; and it will be 1 or -1 for other situations. The threshold 2 can be obtained by adding a factor on the basis of the threshold 1, that is: in which, the "a" is a number bigger than 1 and the optimum value of "a" can be obtained by emulation and real field test (when the value of "a" is taken as infinite, it is equivalent that only the threshold 1 is used). Fig. 7 is a deciding process of using the double thresholds. The decision principle is mentioned above, that is, when the energy of the bit to be decided is less than Tl, it is decided as 0; otherwise, when the energy of the bit to be decided is stronger than T2, it is also decided as 0; in other situations it is decided as 1 or -1. The method of double-threshold can effectively reduce the errors in the decision process, avoid the error accumulation during multistage process, and reduce the generation of the added interference. In a CDMA system, the reverse pilot channel is sent together with the traffic channel. They suffer same wireless channel fading, so the form of energy envelope of the pilot channel is basically same as that of the traffic channel. When the traffic channel energy becomes stronger, the energy of the pilot channel becomes stronger too. Thus, it can be figured out that the envelope form of the decision threshold is basically same as that of the traffic channel, because the decision threshold according to the above method is generated by using the signal of the pilot estimator. In other words, the envelope form of the decision threshold is basically same as the energy envelop form of the bit stream to be decided. In this way, the stronger the bit stream energy, the bigger the value of the decision threshold, so the more possible the bit stream is decided as 0. According to the above description, when the energy of the bit stream to be decided is stronger, the confidence of its decision is higher; and on the contrary, the confidence is lower when the energy of the bit stream is weak. Since the energy change of the bit stream is continuous and gradual, an idea could be brought forward that at the period that the bit stream energy is stronger, the probability that the bit is decided as 0 should be reduced, while at the period that the bit stream energy is weaker, the probability that the bit is decided as 0 should be maintained or even increased. In such way, the precision of the decision is further improved. To follow the above discussion, a confidence factor can be added to the decision threshold. In the structure, a pilot estimator C can be added into the ICU. The output of the pilot estimator C is a long term average of the energy of the pilot channel, which is used for measuring the output energy of the pilot estimator B. The realizing structure of the pilot estimator C is different from that of the pilot estimators A and B, expressed as: It is a long term average of the pilot channel energy from the time establishing channel to now. The reason for using the long term average is that it is relatively stable and less affected by the multi-path fading, and the energy fluctuation is also mild, so it is suitable to become a measurement reference of the output signal of the pilot estimator B. The measurement reference is obtained by an equation below: Bas = b-EC in which the "b" is a factor less than 1 and bigger than 0, and its optimum value is determined by emulation and real field test. The E_C is an output value of the pilot C. Then, the equation (1) for deciding the threshold can be modified as: in which the K3 is a confidence factor not less than 1 , and its adjustment process is shown in Fig. 8. The output signal of the pilot estimator C is obtained firstly. And next the measuring base "Bas" for E_B is calculated according to the equation (5). Then the E_B is compared with Bas. If E_B is bigger than Bas, which shows that the bit stream energy is stronger, the K3 should be reduced. Reducing K3 is equivalent to reducing the decision threshold, and is equivalent to reducing the probability that the bit stream to be decided is decided as 0. When K3 is reduced, it should be monitored whether the K3 has been reduced down to the value less than 1 . If K3 is less than 1 , K3 should be set as 1 . If E_B is less than Bas, which shows that the bit stream energy is weaker, the K3 should be maintained or even increased. Increasing K3 is equivalent to increasing the decision threshold, and is equivalent to increasing the probability that the bit stream to be decided is decided as 0. When K3 is increased, it should be monitored whether the K3 has been increased up to a value larger than the maximum MAXJG. If true, K3 should be set as MAX_K3. The MAX_K3 is an integer bigger than 1. Its value can be determined by emulation and real field tests. In the whole technology concerned the interference cancellation, precisely reconstructing useful signals of every user is a key technology for the interference cancellation, in which precisely deciding on the de-spread bit stream in ICU is a base for precisely reconstructing useful signals. Regarding how to reduce error rate of the decision, the present invention provides a serial of methods. Compared with other methods, the methods of the invention are more effective and easy to realize. According to the methods of the present invention, the input information needed is easy to obtain, the demodulation performance of a receiver is obviously improved, and it facilitates to increase the system capability and coverage radius. The method and unit for the interference cancellation according to the present invention are mainly used for a receiver with a multi-user detection device so as to increase the demodulation performance and system capability. The method and unit for the interference cancellation according to the present invention applies to not only the parallel interference cancellation structures, but also the serial interference cancellation structures without more modification. The variation of the present invention can also be used for other digital communication systems and analog communication systems that have a mobile station transmitting a pilot channel according with the standard IS-665. INDUSTRIAL PRACTICAL APPLICABILITY Compared with the traditional technologies, the present invention overcomes the disadvantages such as that when it is less than the threshold the reliability of the received signal is worse and decision error or incorrect decision is likely to occur; and during multistage process the performance reduces rapidly caused by accumulated errors. In the present invention signal decision and recovery are more precisely controlled and processed, the precision of the interference cancellation is increased, and therefore the system performance is improved. Furthermore, in the present invention the system capability is increased and the influence of "near-far effect" for the system performance is reduced. Using a receiver with multi-user detection structured by ICU of the present invention can increase the demodulation performance of a base station and system capability. We Claim, 1. A multi-user interference cancellation unit (ICU), said ICU comprising: a de-spread unit (302) for de-spreading and separating useful signal from total input signal; characterizing a plurality of pilot estimators A (304), B (308) and C (310) for receiving de-spread signal from said de-spread unit (302) wherein said pilot estimators B and C generate threshold for signal deciding to cancel interference; a first conjugate device (306) for conjugating the output signal of the pilot estimators; a first multiplication device (312) for multiplying the de-spread bit stream by the signal conjugated through said conjugate device(306); a decision device (316) for deciding the bit stream from said first multiplication device (312) by using the output signals from the pilot estimators B (308) and C ( 310) and system information; the second multiplication device (318) for multiplying the signal output from the decision device (316) by the signal output from the pilot estimator A (304); and the resuU of multiplying is an input of subsequent stage of a demodulation module . 2. A method for using a multi-user interference cancellation unit (ICU) as claimed in claim in 1 ,wherein the method comprises the following steps: a) implementing an operation of de-complex spread for input base band signal by using a de-spread unit b) implementing Walsh de-spread of channels for the de-spread signal to separate the channels of a user; c) multiplying bit streams of each separated channel by a conjugate signal of an output estimate value of a pilot estimator A to cancel the effect of multi-path fading d) sending the bit streams multiplied to a decision device for signal deciding; e) constructing a threshold for signal deciding by using other two pilot estimators B and C and the number of users, channel types and service types provided by the system, and then deciding on the input bit streams; and f) multiplying the bit streams decided by an estimate value output by the pilot estimator A to reconstruct the effect of multi-path fading and subsequent reconstruction of the signal. 3. The method as claimed in claim 2, wherein, the step d, comprises : dl) obtaining parameters such as the number of the users, and channel types or service types from the system; d2) determining the information of channel types or service types; d3) adjusting the coefficient K2 as herein described according to the information of the step d2; d4) determining the number of users in the current system; and d5) adjusting the coefficient Kl in step d3 according to the number of users. 4. The method as claimed in claim 3, wherein in the step d3, higher value of K2 is taken , the value of K2 corresponding with convolution code channel is less than that corresponding with the channel of Turbo code type. 5. The method as claimed in claim 4, wherein, when adjusting the coefficients, K2 is determined prior to Kl, the value of K2 increases linearly with increase of channel rate and the change of channel code type, generally the value of Kl is set from 2 to 3, in order to simulate and test, the optimal value of K2 under different channel rates and different code types. 6. The method as claimed in claim 2, wherein in the step e, a double thresholds decision is adopted for deciding input bit stream, when the energy of a bit to be decided is less than threshold 1 (Tl) or is bigger than the threshold 2 (T2), the value "0" is decided; and for other situations , the value "1" or "-1" is decided respectively.. 7. The method as claimed in claim 3, wherein in the step d3, the confidence factor coefficient K3 is not less than 1. 8. The method as claimed in claim 7, wherein steps of step d3 comprises: a) obtaining the output signal of the pilot estimator; b) calculating the measuring base "Bas" for E_B, as herein described, and c) adjusting the values of K3 accordingly as herein described; 9. A multi-user interference cancellation unit (ICU) substantially as herein described with reference to the accompanying specification, examples and drawings. 10. A method for multi-user interference cancellation substantially as herein described with reference to the accompanying specification, examples and drawings.

Full Text

A Method and Unit for Multi-User Interference Cancellation
FIELD OF THE INVENTION
The invention relates to multi-user detection in a code division multiple access (CDMA) communication system, more particularly, to a method and unit of multi-user interference cancellation.
BACKGROUND OF THE INVENTION
CDMA is one of multiple-access modulation technologies widely used in mobile communication. Other multiple-access modulation technologies include Time Division Multiple Access (TDMA) and Frequency Division Multiple Access (FDMA). CDMA technology has advantages of high frequency band utilization and large system capacity over other multiple modulation technologies.
In CDMA communication system, signals of each user overlap in time domain and frequency domain, which are distinguished only by their spreading codes. If these spreading codes are completely orthogonal to each other, the signals of each user can be completely recovered by using a correlator and matched filter. In practical system, however, the users are not synchronized and the signals of various users arrive at the receiver with various time delays, therefore it is really hard to find out a kind of spreading code sequences that can make the signals of all the users to be orthogonal within all possible relative time delays. Since spreading code sets are not completely orthogonal, there are interferences among various users or various multi-paths of one user, which is so-called multiple access interference (or multi-user interference). A conventional receiver demodulates the signals by using a correlator (or matched filter), but the interference among users could not be cancelled during a correlating process. For a certain user, the signals from other users are all deemed as noise. Therefore, with the increase of users in the system, the interferences among the users increase gradually. When the interferences are cumulated to a certain degree and exceed the minimum SNR required in system modulation, the system is unable to allow more users to access. Therefore, CDMA system is a system of interference limited. To solve the problem of interference limited in a CDMA system, the influence from multi-user interference must be reduced. In fact, unlike system thermal noise, multiple access interference (MAI) can be estimated and regenerated theoretically. Therefore it is possible to reduce the MAI in signals received by integrating useful information of every user and adopting certain signal processing methods, which is the object to be realized in the

multi-user detection technology. Effective multi-user detection technology can increase system capacity and system covering radius, and alleviate the problem of interference limited in a CDMA system.
Meanwhile, interference cancellation (multi-user detection) technology can effectively reduce the influence of "near-far effect" on the system performance. Because of the influence of wireless channel fading and the difference of the distances between a base station and mobile stations, the signal power of each mobile station received by the base station is different. A user with strong power signal will impose strong interference on a user with weak power signal, so that the performance of the user with weak power signal is reduced, and even may not operate normally. Using power control technology can make almost equal power among all the mobile stations received by base station, and alleviate the "near-far effect" to a certain degree. However, power control still has some disadvantages, such as, using channel to transmit power control information, control lag, performance related to the rate of mobile users, etc. Furthermore, power control could not solve the problem of system capacity limited by MAI. Power control aims at limiting the interference to an acceptable level. Unlike power control, interference cancellation technology is used to eliminate the interferences among the users to a maximum extent, so as to eliminate the cause of "near-far effect" radically. Therefore, the interference cancellation technology can alleviate effectively the influence of "near-far effect", and relieve the performance requirement for power control in the system.
Generally, mobile communication faces a time-varying multi-path fading environment. In this environment, the signals transmitted through various paths will arrive at a base station with various time delays. Since the spreading codes are not completely orthogonal, mutual interferences similar to MAI exist among signals of various paths of the same user. In design of a conventional CDMA receiver, Rake branch is adopted to demodulate respectively signals of several paths with the strongest power for each user, and then maximal ratio combination is implemented, so that the effect of multi-path fading is overcome by using diversity reception technique. During demodulating for each path, however, other multi-paths are still deemed as noise, and information containing thereof could not be used. Thus, in the environment of multi-path propagation, multi-access interference and multi-path influence could not be overcome simultaneously only by simple diversity reception and combination. If the multi-path influence is considered in the algorithm of multi-user detection, it is obvious that the effective information can be adequately used and the system performance will be further improved.

A multi-user detection receiver can be classified as linear multi-user receiver and nonlinear multi-user receiver depending on their structure with or without feedback.
A linear multi-user detector equates MAI in multi-user communication environment with a transmission response matrix of a channel. The transmission matrix is related with spread sequence of each user and the relative time delay between spread sequences of various users. If an inverse matrix of a channel transmission matrix is obtained, multi-user signals can be output through K (K is the number of users) matched filters, and implemented inverse operation by using the matrix, so that the correlation among users can be cancelled equivalently and the object of canceling MAI is achieved. However, in this method, exact phase information between spread codes should be known, and the inverse matrix of related matrix should be calculated momentarily according to any change of practical users or environment. Hence, the algorithm of the method is complicated, the quantity of calculation is great, and real time realization is not easy to approach.
The other kind of important multi-user detector is called nonlinear multi-user detector (also called subtract!ve interference cancellation detector). In its basic principle, MAI information from each user is estimated independently in a receiving end, next partial or entire MAI of the corresponding user is subtracted from total receiving signal so as to obtain the interference cancelled signal of each user, and then a traditional receiver is used for demodulation. Generally, the feedback-based detector is realized by multistage mode, and expected that interference is cancelled to the maximum extent by using multistage feedback so as to obtain better demodulation performance. From the analysis on a linear multi-user detector, it can be seen that the linear multi-user detector uses matrix operation with complex calculation, which does not facilitate hardware realization. From the view of realizing, nonlinear multi-user detector is more effective.
Nonlinear multi-user detector can be realized by a serial or parallel structure. A detector with a serial structure generally requires firstly ranking input signals according to power, implements interference cancellation to the user with stronger power orderly, takes the signal of interference cancelled as input, and then implemented the same processes to the users with weaker power. Under the situation that the effect of power control is not obvious or is lagged, the performance of serial process is better than that of parallel process, but it will incur process time delay in proportion to the number of users. When the powers of users are relatively balanced among one another or the request of process time delay is higher in the system, the parallel structure is generally adopted. No matter which structure is adopted, multistage processing is commonly implemented to obtain better effect of interference

cancellation.
No matter whether a serial or parallel structure is adopted, the performance of a multi-user detector is determined by an interference cancellation unit (ICU). ICU generally includes two parts of signal deciding and signal recovering. The object of signal deciding and signal recovering is precisely reconstructing the data of each user at the receiving end so as to cancel interference from other users during interference cancellation and meanwhile to ensure that additional interference will not be incurred from errors in signal reconstructing during interference cancellation. The precision of signal deciding and signal recovering will directly affect on signal reconstruction capability of ICU, which will determine the integrated performance of the detector. Thus, how to implement exactly signal deciding and signal recovering becomes a critical factor for improving detector's performance.
At present, the research on multistage interference cancellation is generally classified as non-decision interference cancellation and decision interference cancellation (also called soft-decision interference cancellation and hard-decision interference cancellation). Non-decision interference cancellation uses directly output of a correlation receiver to generate interference recovery signal, during which it does no need to estimate channel parameters. Its algorithm is relatively simple and it can obtain the gain of interference cancellation in white noise channel. However, if the effect of multi-path fading is not considered in Rayleigh channel, the signal obtained without through Rake process can not overcome the effect of fading. In addition, during interference recovery if the multi-path signal passed through channel is not reconstructed, the recovery signal obtained by such process is very different with real receiving signal, which is likely to induce interference after interference cancellation and directly results in reducing the system performance under the Rayleigh environment. Decision interference cancellation implements decision by using the signal Rake-combined and recovers multi-path signal during reconstruction so as to cancel multi-user interference and multi-path interference effectively. According to patent 6081516 of NEC, "Multiuser Receiving Device for Use in A CDMA System", firstly the hard-decision is executed for the data Rake-combined, the multi-path signal is recovered respectively by using path estimation, and then interference cancellation is implemented, which is applicable to a Rayleigh fading environment. However, because the decision is implemented without corresponding process based on the characteristics of signal received, when the received signal amplitude of a certain user or path is less, the reliability of implementing decision and recovery by using the signal is relatively lower. Under this situation, since the interference recovery signal is inaccurate, the interference is induced

during interference cancellation.
In "Third Generation Mobile Radio Systems Using Wideband CDMA Technology and Interference Canceller for Its Base Station" (see FUJITSU Sci. Tech. J., 34, 1, pp.50-57 (September 1998)), some technologies are adopted to improve the precision of signal decision and signal recovery. A mixed decision is firstly implemented at signal decision. Particularly, a threshold is set according to a energy of path estimate, if the energy of the signal Rake-combined is bigger than the threshold, +1 or -1 is decided; if the energy of the signal Rake-combined is less than the threshold, a value (less than 1) normalized by the threshold is decided. It is understandable that the reliability of received signal is relatively higher when it is bigger than the threshold, and the decision can be deemed precise; when it is less than the threshold, however, the reliability of received signal is worse, and the decision is likely to be an error. If the decision is wrong, the interference will on the contrary be increased during interference cancellation even though the result of decision is a relatively smaller amplitude value. In this way, the accumulation of errors will reduce rapidly the performance during such multistage process.
In "Successive Interference Cancellation for Multiuser Asynchronous DS/CDMA Detectors in Multipath Fading Links" (see IEEE TRANSACTIONS ON COMMUNICATIONS, VOL.46, NO.3, MARCH 1998), the method of threshold decision is also adopted. Unlike the above paper, the decision is not implemented when the energy of the combined signal is less than the threshold. In addition, the threshold is calculated according to the energy variance of the received signals from users to be demodulated, the energy variance of users interfering, and energy variance of noise. But in the practical system, it is not a simple process to obtain the precision value of three energy variances. A more complex calculation is needed to determine the threshold related with three parameters according to changes of an environment.
SUMMARY OF THE INVENTION
The object of the present invention is to provide a method and unit for multi-user interference cancellation to overcome the disadvantages of inaccurate signal decision and signal recovery existing in the current interference cancellation unit.
To achieve the above object, the present invention adopts the following technical solution.
A method for multi-user interference cancellation according to the present invention includes the steps as follows:

a) implementing an operation of de-complex spread for input base band signal by using a de-spread unit
b) implementing Walsh de-spread of channels for the de-spread signal to separate the channels of a user;
c) multiplying bit streams of each separated channel by a conjugate signal of an output estimate value of a pilot estimator A to cancel the effect of multi-path fading;
d) sending the bit streams multiplied to a decision device for deciding;
e) constructing a threshold for deciding by using other two pilot estimators B and C and the number of users, channel types and service types provided by system, and then deciding the input bit streams; and
f) multiplying the bit streams decided by an estimate value output by the pilot estimator A to reconstruct the effect of multi-path fading and subsequent reconstruction of the signal.
In the above step d, when deciding, the following steps are further included:
dl) obtaining parameters such as the number of the users, and channel types or service types from the system;
d2) determining the information of channel types or service types;
d3) adjusting the coefficient K2 in the equation of Threshold = K1.K2.E_B according to the information of the step d2;
d4) determining the number of users in the current system; and
d5) adjusting the coefficient Kl in the equation in step d3 according to the number of users.
In the step d3, when adjusting the coefficient K2, the higher channel rate, the higher the value of K2 is taken. The value of K2 corresponding with convolution code channel should be less than that corresponding with the channel of Turbo code type.
When adjusting the coefficients, K2 is determined first, and then Kl is determined. The value of K2 is approximately linear increase with an increase of channel rate and the change of channel code type. In practical operation, Kl can be set provisionally as about 2 to 3, and then to simulate and test the optimal value of K2 under different channel rates and different code types.
In the step e, when deciding, a double thresholds decision can be adopted. Namely, when the energy of a bit to be decided is less than threshold 1 (Tl), the value of "0" is decided; when the energy of the bit to be decided is bigger than the threshold 2 (T2), the value of "0" is decided too; the value of "1" or "-1" is decided for other situations.

In the step d3, the adjustment also can be implemented according to the equation of
When adjusting, first the output signal of the pilot estimator C is obtained; next the measuring base "Bas" for E_B is calculated according to the equation of
whether the E_B is bigger than Bas is judged, if true, it is shown that the bit energy is stronger and the value of K3 should be reduced; then whether K3 has been reduced down to the value less than 1 is judged, if true, K3 = 1;
if E_B is less than Bas, it is shown that the bit energy is weaker and the value of K3 should be held or increased;
and then whether K3 has exceeded a maximum MAX_K3 is judged, in which MAX_K3 is an integer bigger than 1 .
A multi-user interference cancellation unit of the present invention includes a de-spread unit, three pilot estimators A, B and C, a first multiplication device, a conjugate device, a decision device, and a second multiplication device; the de-spread unit de-spreads an input signal, and the useful signal of the user is separated from total signal; on the one hand, the de-spread signal is sent to the three pilot estimators, and on the other hand, the de-spread bit stream is sent to the first multiplication device; the output signal of the pilot estimator A is conjugated by the conjugate device; the first multiplication device multiplies the de-spread bit stream by the signal conjugated through the conjugate device; the decision device uses the output signals of the pilot estimators B and C and system information to decide the bit stream to be decided that is output from the first multiplication device; the second multiplication device multiplies the signal output from the decision device by the signal output from the pilot estimator A; the result of multiplying is an input of subsequent stage of a demodulation module.
BRIEF DESCRIPTION OF THE DRAWINGS
Fig. 1 is a schematic diagram of a traditional receiver;
Fig. 2 is a structure schematic diagram of traditional parallel interference cancellation for multi-user detection;
Fig. 3 is a structure schematic diagram of an interference cancellation unit according to the present invention;

Fig. 4 is a flow chart of an interference cancellation method according to the present invention;
Fig. 5 is a process schematic diagram of an interference cancellation method (changing decision threshold according to different situations) of the present invention;
Fig. 6 is a demodulation performance curve of a receiver under different decision thresholds according to the method of the present invention;
Fig.7 is a process schematic diagram of decision (how to reduce the error rate of decision) in an interference cancellation method of the present invention;
Fig.8 is a process schematic diagram of decision (how to determine the confidence of decision) in an interference cancellation method of the present invention.
DESCRIPTION OF THE PREFERRED EMBODIMENT(S)
Fig. 1 shows a basic principle of a receiver. A RF module converts a RF signal received into a base band signal, and sends the base band signal to a demodulation module for demodulating. The symbol stream demodulated is sent to a decoder for decoding. Generally, a RF processor includes a mixer and a local oscillator, which can convert the RF signal into an intermediate frequency signal, and then further into the base band signal.
The demodulation module includes a Rake receive module for searching K pieces of the strongest multi-path signals; a de-spread module uses hybrid PN spread sequence corresponding to each user to de-spread each path signal of each user; finally, the de-spread symbol stream is multiplied by Walsh function sequence corresponding to each channel, and then the bit stream obtained by the multiplying is sent to the decoder for decoding.
The decoder chooses related decoding mode according to different coding mode of signals. For example, the fundamental channel uses convolution coding mode, the corresponding decoding mode is Viterbi decoding; while the supplemental channel uses Turbo coding mode, the corresponding decoding mode is Turbo decoding.
Fig. 2 shows a parallel interference cancellation (PIC) structure for a traditional multi-user detection. The structure includes M-stages multi-user detection process (M>1). Each stage includes N pieces of ICUs, in which N is the number of users. The N ICUs adopt the parallel arrangement mode and implement the interference cancellation simultaneously, unlike in serial mode interference cancellation in which the strongest user is chosen to implement interference cancellation firstly. The delay length of the delay units 201-1 to 201-M of each stage equals to the process time delay of the corresponding stage, to ensure that the two input signals of the subtracter 204-n (n is 1 ~ M) are synchronous.

It is should be noted that the interference cancellation structures of the first stage and the last stage have little difference from that of other stages. The input signal of the first stage is taken from the output of the matched filter 206, while the input signals of other stages are all taken from the output of the subtracter of the preceding stage, namely the signals that the interference cancellation has been implemented. The ICU 2-M-l to 2-M-N of the last stage only need to implement interference signal reconstruction. After the multi-user detection of all stages is finished, the final output signal An is an interference reconstruction signal by subtracting all other users except user n from total signal (n is 1~N). In theory, if the signals of all users can be reconstructed accurately, the demodulation performance even in a multi-user situation is same as that of one user.
Fig. 3 is a structure schematic diagram of ICU in Fig.2, namely an ICU structure adopted by the present invention. The unit includes a de-spread unit, three pilot estimators A, B and C, a first multiplication device, a conjugate device, a decision device, and a second multiplication device, the de-spread unit de-spreads an input signal, and separates the useful signal of the user from total signal, on the one hand, the de-spread signal is sent to three pilot estimators, on the other hand, the de-spread bit stream is sent to the first multiplication device, the output signal of pilot estimator A is conjugated by the conjugate device, the first multiplication device multiplies the de-spread bit stream by the signal conjugated through the conjugate device; the decision device uses the output signals of pilot estimators B and C and system information to decide the bit stream to be decided output from the first multiplication device; the second multiplication device multiplies the signal output from the decision device by the signal output from the pilot estimator A, the result of multiplying is an input of demodulation module of the subsequent stage.
In a traditional ICU structure, only one pilot estimator is used. The signal of the pilot estimator is used not only for weighting the de-spread signal and reconstructing the received signal, but also as a reference signal of the decision device. However, the essential of weighting the de-spread signal and reconstructing the received signal is to cancel and reconstruct the multi-path fading effect in the received signal, which requests that the pilot estimator should not have an overlong cumulated length and time delay so that it can comprise enough exact multi-path fading components, but this kind of pilot estimate is not applicable to the reference signal of a decision device. Thus, in the ICU structure provided by the present invention, special pilot estimators B and C are added to provide the reference signal of a decision device, while the pilot estimator A is used specially for weighting the de-spread signal and reconstructing the received signal.

Fig. 4 is a flow chart of interference cancellation method according to the present invention. In the method of the present invention, in the first step 401, the input signal is de-spread by using the hybrid PN spread sequence of the corresponding user, the useful signal of the user is separated from the total signal, and the de-spread signal is then sent to the pilot estimators A~C. In the second step 402, the Walsh de-spread is implemented for the PN de-spread signal to separate the signals of each channel of each user. The steps 401 and 402 are implemented within the de-spread unit. The steps 404, 408 and 410 extract the pilot estimates A, B and C according to different pilot estimate mode, respectively. In the step 412, the bit stream of each channel separated is multiplied by the conjugate signal of output signal of the pilot estimator A to cancel the effect of multi-path fading, and the bit stream multiplied is sent to the decision device for deciding. In the step 416, a threshold for deciding is constructed by using the other two pilot estimators B and C and the system information provided by step 414, and then the input bit streams to be decided is decided. In the step 418, the bit stream decided is multiplied by the pilot estimate A to reconstruct the effect of multi-path fading, and then the subsequent reconstruction processes of the signal are implemented, such as Walsh spread, hybrid PN sequence spread, which are not described in details here. In the above method, the output of the pilot estimators B and C and system information are all inputs needed for deciding by the decision device, and the detail function of the device will be described below.
The function of the decision device is to decide the input bit stream as a sequence of 1 and -1. The most simply decision method is that if the amplitude of a certain bit is bigger than level 0, it is decided as 1, otherwise as -1. This decision method is very inaccurate if considering the effect of thermal noise and multi-path fading. In order to decide more accurately and avoiding additional interference incurred as much as possible, it is necessary for the decision process to introduce a threshold. The new decision method is that if the amplitude of a certain bit is bigger than the threshold, it is decided as 1; if the amplitude is less than the negative threshold, it is decided as -1; level 0 is decided for other situations. In this method, it is considered that the decision confidence of the bit is lower when the bit amplitude is weaker, it is better to decide it as 0 to avoid possible additional interference incurred, rather than a non-zero value. In addition, because of a wild fluctuation of bit stream energy caused by multi-path fading, using a fixed decision threshold will obviously not yield an exact decision result. Meanwhile, it is difficult to predict a value of the bit energy, so a fixed threshold is really hard to realize. Considering the above factors, the threshold of the present invention can be obtained by an equation below:

Threshold = K1.K2.E_B (1)
in which, E_B is an energy value of output signal of the pilot estimator B, the value of Kl depends on the number of users, the value of K2 depends on the channel type or service type of a corresponding user, and Kl and K2 are non-negative numbers. For example, if the user uses a supplemental channel of 153.6 kbps, its K2 is bigger than a fundamental channel of 9.6 kbps. Kl and K2 are obtained by emulation and real field test. Fig.6 is a curve diagram of demodulation performance obtained by emulation under different decision thresholds of different user numbers, and the channel used is the fundamental channel of 9.6 kbps. It can be seen that the best decision threshold is different for the different number of users. It means that the change of Kl follows the change number of users. The optimum Kl and K2 could be concluded based on a mass of emulation data and real field test data.
Fig.5 is a process schematic diagram of interference cancellation method (changing decision threshold according to different situations) according to the present invention.
In this process, in the first step 502, the parameters, such as the number of the users and channel type or service type, are obtained from system, and it is easy to obtain the parameters. In the second step 503, the channel type or service type is determined, for example, it is a fundamental channel or supplemental channel, what is the channel rate, and which coding type is adopted. In the third step 504, the coefficient K2 is adjusted according to the information of the step 503, for example, the higher the channel rate, the higher the value of K2; the value of K2 related to the channel of convolution code type is less than that related to the channel of Turbo code type. In general, the value of K2 tends to approximately linearly increase with an increase of channel rate and the change of code type (changing from convolution code to Turbo code). In a practical operation, Kl can be set as from about 2 to about 3 provisionally, and emulate and test the optimum value of K2 under different channel rate and different code type. Because the optimum value does not much relates to number of users, the number of users for testing can be 10 to 40. After the K2 is determined, in the fourth step 505, the current number of users is determined. In the fifth step 506, Kl is adjusted according to the number of users. For example, referring to Fig.6, when the number of users ranges from 10 to 20, the optimum Kl is 1; when the number of users is 30, the optimum Kl is 2; when the number of users is 40, the optimum Kl is 4 or 5. It is easy to determine the range of Kl by emulating, which should be not too big or too small. Fig.6 is an emulation result only under a special environment. A mass of emulation and real field test should be done to obtain the optimum Kl adapted to various environments. Although the
optimum Kl may offset in different communication environment, the offset is not too big. Therefore, a balance point could be chosen among the optimum values to basically satisfy various environments. Such balance point is the optimum value of Kl adapted to the number of current users. In practical realizing, a table of Kl and K2 could be made according to testing data, in which the system can query the related Kl and K2 according to real system information (the parameters such as the number of uses, the channel type, code type).
The E_B in the above equation (1) uses an energy value of the pilot B. The traditional ICU has only one pilot estimator A. Because the pilot estimator needs to weight the bit stream before and after decision to cancel and reconstruct the effect of multi-path fading, the time delay and accumulated length of the pilot estimator should not be overlong. However, the variance of the signal output from such pilot estimator is bigger and the energy fluctuation is wilder, which is not suitable to generate the decision threshold. Therefore the pilot estimator B should be added to do the work. The difference between the pilot estimator B and A is that the accumulated length of the pilot estimator B is longer, then variance of the output signal is smaller and energy fluctuation is less. Additionally the envelope of its energy satisfies basically the energy envelope of the bit stream to be decided, so it is suitable to generate the decision threshold. From the emulation test, it can be seen that compared the effect of using only the pilot estimator A with using the combining pilot estimator A and B, the demodulation effect is improved obviously by using the pilot estimator B to generate the decision threshold.
When deciding, if the bit energy is weaker, the decision confidence is lower. With the same principle, if the bit energy is too strong over a certain limitation, the decision confidence of the bit is also obviously lower because the energy will be deemed as an abnormal energy resulted from interference. Based on the above discussion, the present invention adopts double-threshold decision to further reduce the additional interference caused by wrong decisions. In the solution of the present invention, when the energy of the bit to be decided is less than the threshold 1 or bigger than the threshold 2, the bit should be decided as 0; and it will be 1 or -1 for other situations. The threshold 2 can be obtained by adding a factor on the basis of the threshold 1, that is:
in which, the "a" is a number bigger than 1 and the optimum value of "a" can be obtained by

emulation and real field test (when the value of "a" is taken as infinite, it is equivalent that only the threshold 1 is used). Fig. 7 is a deciding process of using the double thresholds. The decision principle is mentioned above, that is, when the energy of the bit to be decided is less than Tl, it is decided as 0; otherwise, when the energy of the bit to be decided is stronger than T2, it is also decided as 0; in other situations it is decided as 1 or -1. The method of double-threshold can effectively reduce the errors in the decision process, avoid the error accumulation during multistage process, and reduce the generation of the added interference.
In a CDMA system, the reverse pilot channel is sent together with the traffic channel. They suffer same wireless channel fading, so the form of energy envelope of the pilot channel is basically same as that of the traffic channel. When the traffic channel energy becomes stronger, the energy of the pilot channel becomes stronger too. Thus, it can be figured out that the envelope form of the decision threshold is basically same as that of the traffic channel, because the decision threshold according to the above method is generated by using the signal of the pilot estimator. In other words, the envelope form of the decision threshold is basically same as the energy envelop form of the bit stream to be decided. In this way, the stronger the bit stream energy, the bigger the value of the decision threshold, so the more possible the bit stream is decided as 0. According to the above description, when the energy of the bit stream to be decided is stronger, the confidence of its decision is higher; and on the contrary, the confidence is lower when the energy of the bit stream is weak. Since the energy change of the bit stream is continuous and gradual, an idea could be brought forward that at the period that the bit stream energy is stronger, the probability that the bit is decided as 0 should be reduced, while at the period that the bit stream energy is weaker, the probability that the bit is decided as 0 should be maintained or even increased. In such way, the precision of the decision is further improved.
To follow the above discussion, a confidence factor can be added to the decision threshold. In the structure, a pilot estimator C can be added into the ICU. The output of the pilot estimator C is a long term average of the energy of the pilot channel, which is used for measuring the output energy of the pilot estimator B. The realizing structure of the pilot estimator C is different from that of the pilot estimators A and B, expressed as:
It is a long term average of the pilot channel energy from the time establishing channel to now. The reason for using the long term average is that it is relatively stable and less

affected by the multi-path fading, and the energy fluctuation is also mild, so it is suitable to become a measurement reference of the output signal of the pilot estimator B. The measurement reference is obtained by an equation below: Bas = b-EC
in which the "b" is a factor less than 1 and bigger than 0, and its optimum value is
determined by emulation and real field test. The E_C is an output value of the pilot C.
Then, the equation (1) for deciding the threshold can be modified as:

in which the K3 is a confidence factor not less than 1 , and its adjustment process is shown in Fig. 8. The output signal of the pilot estimator C is obtained firstly. And next the measuring base "Bas" for E_B is calculated according to the equation (5). Then the E_B is compared with Bas. If E_B is bigger than Bas, which shows that the bit stream energy is stronger, the K3 should be reduced. Reducing K3 is equivalent to reducing the decision threshold, and is equivalent to reducing the probability that the bit stream to be decided is decided as 0. When K3 is reduced, it should be monitored whether the K3 has been reduced down to the value less than 1 . If K3 is less than 1 , K3 should be set as 1 . If E_B is less than Bas, which shows that the bit stream energy is weaker, the K3 should be maintained or even increased. Increasing K3 is equivalent to increasing the decision threshold, and is equivalent to increasing the probability that the bit stream to be decided is decided as 0. When K3 is increased, it should be monitored whether the K3 has been increased up to a value larger than the maximum MAXJG. If true, K3 should be set as MAX_K3. The MAX_K3 is an integer bigger than 1. Its value can be determined by emulation and real field tests.
In the whole technology concerned the interference cancellation, precisely reconstructing useful signals of every user is a key technology for the interference cancellation, in which precisely deciding on the de-spread bit stream in ICU is a base for precisely reconstructing useful signals. Regarding how to reduce error rate of the decision, the present invention provides a serial of methods. Compared with other methods, the methods of the invention are more effective and easy to realize. According to the methods of the present invention, the input information needed is easy to obtain, the demodulation performance of a receiver is obviously improved, and it facilitates to increase the system capability and coverage radius.
The method and unit for the interference cancellation according to the present invention are mainly used for a receiver with a multi-user detection device so as to increase the

demodulation performance and system capability.
The method and unit for the interference cancellation according to the present invention applies to not only the parallel interference cancellation structures, but also the serial interference cancellation structures without more modification. The variation of the present invention can also be used for other digital communication systems and analog communication systems that have a mobile station transmitting a pilot channel according with the standard IS-665.
INDUSTRIAL PRACTICAL APPLICABILITY
Compared with the traditional technologies, the present invention overcomes the disadvantages such as that when it is less than the threshold the reliability of the received signal is worse and decision error or incorrect decision is likely to occur; and during multistage process the performance reduces rapidly caused by accumulated errors. In the present invention signal decision and recovery are more precisely controlled and processed, the precision of the interference cancellation is increased, and therefore the system performance is improved. Furthermore, in the present invention the system capability is increased and the influence of "near-far effect" for the system performance is reduced. Using a receiver with multi-user detection structured by ICU of the present invention can increase the demodulation performance of a base station and system capability.

We Claim,
1. A multi-user interference cancellation unit (ICU), said ICU comprising:
a de-spread unit (302) for de-spreading and separating useful signal from total input signal; characterizing a plurality of pilot estimators A (304), B (308) and C (310) for receiving de-spread signal from said de-spread unit (302) wherein said pilot estimators B and C generate threshold for signal deciding to cancel interference;
a first conjugate device (306) for conjugating the output signal of the pilot estimators;
a first multiplication device (312) for multiplying the de-spread bit stream by the signal conjugated through said conjugate device(306);
a decision device (316) for deciding the bit stream from said first multiplication device (312) by using the output signals from the pilot estimators B (308) and C ( 310) and system information;
the second multiplication device (318) for multiplying the signal output from the decision device (316) by the signal output from the pilot estimator A (304); and the resuU of multiplying is an input of subsequent stage of a demodulation module .
2. A method for using a multi-user interference cancellation unit (ICU) as
claimed in claim in 1 ,wherein the method comprises the following steps:
a) implementing an operation of de-complex spread for input base band signal by using a de-spread unit
b) implementing Walsh de-spread of channels for the de-spread signal to separate the channels of a user;
c) multiplying bit streams of each separated channel by a conjugate signal of an output estimate value of a pilot estimator A to cancel the effect of multi-path fading
d) sending the bit streams multiplied to a decision device for signal deciding;
e) constructing a threshold for signal deciding by using other two pilot estimators B and C and the number of users, channel types and service types provided by the system, and then deciding on the input bit streams; and
f) multiplying the bit streams decided by an estimate value output by the pilot estimator A to reconstruct the effect of multi-path fading and subsequent reconstruction of the signal.
3. The method as claimed in claim 2, wherein, the step d, comprises :
dl) obtaining parameters such as the number of the users, and channel types or service types from the system;

d2) determining the information of channel types or service types; d3) adjusting the coefficient K2 as herein described according to the information of the step d2;
d4) determining the number of users in the current system; and
d5) adjusting the coefficient Kl in step d3 according to the number of users.
4. The method as claimed in claim 3, wherein in the step d3, higher value of
K2 is taken , the value of K2 corresponding with convolution code channel is
less than that corresponding with the channel of Turbo code type.
5. The method as claimed in claim 4, wherein, when adjusting the coefficients, K2 is determined prior to Kl, the value of K2 increases linearly with increase of channel rate and the change of channel code type, generally the value of Kl is set from 2 to 3, in order to simulate and test, the optimal value of K2 under different channel rates and different code types.
6. The method as claimed in claim 2, wherein in the step e, a double thresholds decision is adopted for deciding input bit stream, when the energy of a bit to be decided is less than threshold 1 (Tl) or is bigger than the threshold 2 (T2), the value "0" is decided; and for other situations , the value "1" or "-1" is decided respectively..
7. The method as claimed in claim 3, wherein in the step d3, the confidence factor coefficient K3 is not less than 1.
8. The method as claimed in claim 7, wherein steps of step d3 comprises:
a) obtaining the output signal of the pilot estimator;
b) calculating the measuring base "Bas" for E_B, as herein described, and
c) adjusting the values of K3 accordingly as herein described;
9. A multi-user interference cancellation unit (ICU) substantially as herein described
with reference to the accompanying specification, examples and drawings.
10. A method for multi-user interference cancellation substantially as herein described
with reference to the accompanying specification, examples and drawings.

Documents:

2783-delnp-2005-abstract.pdf

2783-DELNP-2005-Claims-(11-11-2008).pdf

2783-delnp-2005-claims-08-08-2008.pdf

2783-delnp-2005-claims.pdf

2783-delnp-2005-complete specification (granted).pdf

2783-DELNP-2005-Correspondence-Others-(05-11-2007).pdf

2783-DELNP-2005-Correspondence-Others-(08-08-2008).pdf

2783-delnp-2005-correspondence-others.pdf

2783-DELNP-2005-Description (Complete)-(05-11-2007).pdf

2783-DELNP-2005-Description (Complete)-08-08-2008.pdf

2783-delnp-2005-description (complete).pdf

2783-delnp-2005-drawings.pdf

2783-DELNP-2005-Form-1-(05-11-2007).pdf

2783-delnp-2005-form-13-(08-08-2008).pdf

2783-delnp-2005-form-18.pdf

2783-DELNP-2005-Form-2-(05-11-2007).pdf

2783-delnp-2005-form-2.pdf

2783-delnp-2005-gpa.pdf

« Previous Patent

Next Patent »

Patent Number

225772

Indian Patent Application Number

2783/DELNP/2005

PG Journal Number

01/2009

Publication Date

02-Jan-2009

Grant Date

28-Nov-2008

Date of Filing

22-Jun-2005

Name of Patentee

ZTE CORPORATION

Applicant Address

ZTE BUILDING SOUTH HI-TECH ROAD, HI-TECH INDUSTRIAL PARK, NANSHAN DISTRICT, SHENZHEN, GUANGDONG 518057 (CN)

Inventors:

#	Inventor's Name	Inventor's Address
1	DAI, QIAN	ZTE BUILDING SOUTH HI-TECH ROAD, HI-TECH INDUSTRIAL PARK, NANSHAN DISTRICT, SHENZHEN, GUANGDONG 518057 (CN)
2	LIU, YING	ZTE BUILDING SOUTH HI-TECH ROAD, HI-TECH INDUSTRIAL PARK, NANSHAN DISTRICT, SHENZHEN, GUANGDONG 518057 (CN)
3	ZHOU, MENG	ZTE BUILDING SOUTH HI-TECH ROAD, HI-TECH INDUSTRIAL PARK, NANSHAN DISTRICT, SHENZHEN, GUANGDONG 518057 (CN)

PCT International Classification Number

H04Q 7/32

PCT International Application Number

PCT/CN2002/000859

PCT International Filing date

2002-11-29

PCT Conventions:

#	PCT Application Number	Date of Convention	Priority Country
1	PCT/CN2002/000859	2002-11-29	China