Title of Invention  METHOD AND APPARATUS FOR DETERMINING THE LOCATION OF A MOBILE OBJECT 

Abstract  The invention relates to a positionfinding method for determining the location of a mobile object. The features, for example received field strengths, of a plurality of base stations are measured, and the object position is located from these features, using a reference, map. During an initialization process, a reference map is created with which comprises a multiplicity of positions and the associated featuredependent values. During use of the method, a plurality of positionfinding processes are carried out, by means of each of which a measured featuredependent value and, from this, a located position of the object, are determined using the predetermined reference map. The predetermined reference map is in each case updated for at least some of the positions found, during which updates, the featuredepended values are each corrected by a correction term at the support points of the reference map in a predetermined area surrounding an object position. The correction term depends on the difference between the measured featuredependents value for the respective location, and the featuredependent value at the support point in the predetermined surrounding area. 
Full Text  Description Method and apparatus for determining the position of a mobile object The invention relates to a method and an apparatus for the computeraided localization of a mobile object. Various featurebased methods for finding the position of mobile users or terminal devices in wireless communication networks are known from the prior art. As features, for example, the field strengths of fields through the mobile object are measured, the fields being generated and emitted in the communication network by a plurality of base stations. As communication networks, for example, DECT networks (DECT = Digital Enhanced Cordless Telecommunications), WLAN networks (WLAN = Wireless Local Area Network) and the mobile radiocommunication networks GSM (GSM = Global System for Mobile Communication) and UMTS (UMTS = Universal Mobile Telecommunication System) are used. In featurebased position finding, socalled reference maps are used in which, for a multiplicity of support points, a relationship between the spatial position of the support point and featuredependent variables of the individual base stations is filed. Featuredependent variables in a fieldstrengthbased method are, for example, the field strength of the field emitted by a base station at the spatial position of the support point or the field strength, measured at the base station, of a field that is emitted by an object which is located at the spatial position of the support point. By measuring the field strength at any position and by comparing the measured value with the fieldstrengthdependent variables at the support points on the reference map, the position of the object can be determined. In the fieldstrengthbased method known from the prior art, a calibration or gauging of reference patterns is necessary in order to achieve sufficient accuracy in structured environments. This calibration is labor and costintensive, and the greater the accuracy desired, the more timeconsuming it is. Besides calibration using reference patterns, it is also known from the prior art for the position of the most powerfully receiving base station to be determined in the mobile object by means of simple trilateration models, for example assuming a radial field propagation. In such methods, few parameters are needed, and these can be adjusted by means of few reference measurements. However, these methods have the disadvantage that the accuracy they achieve is significantly lower than in methods in which a reference map is created by means of a multiplicity of measurements. The object of the invention is to create a simple localization method which has high accuracy and can be installed with little outlay and without carrying out a multiplicity of reference measurements. This object is achieved by the independent claims. Further developments of the invention are defined in the dependent claims. In the method according to the invention, the features of a plurality of base stations are measured with the aid of a featurebased positionfinding method and the position of the object located from these features, using a reference map. Here, the feature of a base station is a feature which is associated with a field, the field being a field emitted by the base station at the position of the object or a field emitted by the object at the base station. A relationship is thus created by the field between a single base station and the position of the object in space. Here, the method is initialized in a first step by means of a reference map which comprises a multiplicity of support points, a support point being in each case represented by a position and a featuredependent value assigned to the position. A featuredependent value is understood below to mean a value which depends on the previously described features of the base stations. In particular, the featuredependent value can be a feature vector which comprises the features of all the base stations. In a fieldstrengthbased method, the featuredependent value is, for example, a fieldstrengthdependent value. In particular, the value in such a method can be a fieldstrength vector which comprises as entries the field strengths of each base station. Here, the field strengths of the base stations can be the field strengths of fields emitted by the base stations at the object position. They can, however, also be the field strengths of fields emitted by the object at the location of the individual base stations. The method according to the invention is not restricted to the use of fieldstrengthbased positionfinding methods; in particular, runtimebased or anglebased positionfinding methods can also be used as positionfinding methods. In the case of runtimebased positionfinding methods, the position of an object is determined via the runtime of a field. In the case of anglebased po sitionfinding methods, the position of an object is determined via the spatial angles at which a field strikes an object or a base station. After initialization, during operation a plurality of positionfinding processes are carried out, by means of each of which a measured featuredependent value and, from this, a located position of the object are determined via the featurebased positionfinding method, using the predetermined reference map. One or more updates of the predetermined reference map is/are carried out in each case for at least some of the locations, during which updates the featuredependent values are corrected by a correction term at the support points on the reference map in a predetermined area surrounding an object position, the correction term depending on the difference between the measured featuredependent value of the respective location and the featuredependent value at a support point in the predetermined surrounding area. Here, the predetermined surrounding area can be chosen arbitrarily according to requirements and can, in particular, also cover the entire area of the reference map. Here, the object position about which the predetermined surrounding area is disposed can, for example, be a position known in advance, i.e. it is known in the method at what point the object is located when a corresponding positionfinding process is carried out. In this case, the method according to the invention serves in calibrating a reference map in a highly accurate manner without the method being used for locating unknown positions. The object position can, however, also be an unknown position which is the located position deter mined by means of the respective positionfinding process. In this case, during implementation of the method, a positionfinding process and a calibration, and thus improvement of the positionfinding accuracy, are carried out simultaneously. Here, the updating of the reference map in accordance with the inventive method can always be carried out at the time of the respective positionfinding process, but there is also the possibility for the step not to be carried out until a predetermined number of positionfinding processes have been completed, all object positions up until the preceding positionfinding process then preferably being corrected by a correction term. The reference map that is predetermined during initialization of the method can, for example, be computed by means of a wave propagation model. As the wave propagation model, a radial propagation model, an MWM propagation model (MWM = Multi Wall Model) or a DPM propagation model (DPM = Dominant Path Model) can, for example, be used. These propagation models will be explained in greater detail in the detailed description. Through the use of such wave propagation models, a reference map can be predetermined easily and rapidly during initialization of the method without timeconsuming reference measurements having to be carried out at a multiplicity of positions. It suffices in particular for the parameters of the propagation model to be determined by a simple selfmeasurement by the base stations . In a preferred embodiment, each of the plurality of positionfinding processes is carried out with the aid of a fingerprinting method. The fingerprinting method is adequately known from the prior art and will be explained in greater detail in the detailed description. Optionally, however, other methods can also be used for position finding, for example methods which use a probabilistic cost function based on the differences determined for the correction term in order to determine by this means the located position. In a particularly preferred embodiment of the invention, the size of the correction term decreases in magnitude with increasing distance from the object position, thereby taking account of the fact that the error according to the correction term has only a local influence on the featuredependent value. In a particularly preferred embodiment of the invention, the correction term depends on a Gaussian function with its center in the object position. Furthermore, the predetermined area surrounding the object position also depends on a parameter of the Gaussian function, namely on the standard deviation of the Gaussian function. Here, the predetermined surrounding area is preferably given by a circle about the center of the Gaussian function having a radius which is greater than or equal to the standard deviation and in particular a multiple of, preferably three times, the standard deviation. The values of the Gaussian function outside this circle are so small that they can be ignored. By virtue of the fact that only certain support points in the area surrounding the object position are taken into account, the method requires less computing time. Preferably, the standard deviation of the Gaussian function is chosen such that it depends on the spacing of the support points on the reference map and lies in particular between the supportpoint spacing and five times the supportpoint spacing. In order to reduce the computing time of the method, the Gaussian function is approximated in a further preferred variant of the invention by a function which decreases to zero with increasing distance from the center and, in particular, is zero at a value greater than or equal to the standard deviation, preferably at three times the standard deviation. Such a function is cited for example in the detailed description in equation (10). In a particularly preferred embodiment of the invention, the measured featuredependent value is a feature vector, in particular a fieldstrength vector, measured at the object or at the base stations at the time of the positionfinding process, and the predetermined reference map is a location feature map in which the featuredependent values are feature vectors for the positions of the support points, a feature vector comprising as entries the features of the base stations for the corresponding position. Optionally, the featuredependent value can also be a distance such that the reference map represents a location distance map. Here, a wave propagation model, for example, is used in order to infer from fieldstrength values dominant distances, which are filed in the location distance map. In this case, the positionfinding can be carried out via a trilateration calculation or, where there are more than three base stations, a multilateration calculation. In a particularly preferred embodiment of the invention, the correction term SUPD for a support point k in side the predetermined surrounding area is given by or depends on the following function: where f(r) is a Gaussian function defined as follows: 2l / where a is the standard deviation and r the distance from the object position and K is preferably a value less than 1 and particularly preferably less than or equal to 0.5, where *Ap is a differential feature vector with one entry in each case for the feature of a base station of N base stations, where the differential feature vector kAp is either defined as : *Ap = where °p is the measured feature vector at the time of the positionfinding process and "PRM is a feature vector determined at the located position by means of the location feature map (RM) , or is defined as: *Ap = °p  * where °p is the measured feature vector at the time of the positionfinding process and kpstt is the feature vector at the support point k of the location feature map. According to the cited alternatives for calculating *Ap, either a vector is used as a value for *Ap that is independent of the support point which is corrected, or a value is used in which the value of the feature at the support point under consideration has an influence in accordance with the location feature map. The variant of the method just defined is used for example for carrying out socalled unsupervised learning, in which the position at which the object is located is not known at the time of the respective positionfinding process. This means that the position of the object located by means of the respective position positionfinding process is used as the object position. However, it is also possible for the path of the mobile object to be predetermined and the object position to be known in advance at the time of the respective positionfinding process. Preferably, the object moves on one or more predetermined path sections between, in each case, two base stations, object positions being known in advance on the respective path section Q and a positionfinding process being carried out at each of these object positions. Preferably, the correction term SUPD (r, q) at object position q (q = 1, ..., Q) of Q object positions on a path section is defined as follows: where a is the standard deviation and qr the distance from the object position q and 9r the distance from the object position g and where K is preferably a value less than or equal to 0.5, where £ is a positive term which ensures that the denominator is not zero, where £ is preferably less than 1 and in particular 0.3, where qbp is a differential feature vector comprising one entry in, each case for the feature of a base station of N base stations, where the differential fieldstrength vector qAp is defined as : where cqp is the feature vector measured at the time of the positionfinding process at the object position q and C9ppM is a feature vector determined at the located position by means of the location feature map (RM) . According to this variant, the updating of the location feature map is always carried out after a predetermined path section has been executed, the sum total of all Gaussian functions at the positions on the path section being taken into account in the updating. This variant of the method is a supervised learning method as the object positions are known in advance during the learning method. In a preferred variant of the method just cited, the spacing between consecutive object positions is constant and the standard deviation depends on this constant spacing and is in particular directly proportional to the constant spacing. In a particularly preferred variant of the supervised learning method, the mobile object executes a sequence of predetermined path sections, the order of the path sections executed being suitably determined with the aid of a cost function. The cost function here preferably indicates for a respective path section how great the maximum distance of an object position on the path section is from the nearest base station or from a nearest object position which has already been traversed. By means of such a cost function, it is achieved that a large area of object positions is covered by the method as rapidly as possible. In the method according to the invention, the featurebased positionfinding method can comprise a fieldstrengthbased and/or runtimebased and/or anglebased positionfinding method. Preferably, the runtimebased positionfinding method takes into account only such fields as exhibit a minimum signal strength at the object . The featurebased positionfinding method can use any fields, in particular fields of a DECT and/or WLAN and/or mobile radio network. Besides the inventive method just described, the invention relates furthermore to an apparatus for the computeraided positionfinding of a mobile object with the aid of a featurebased positionfinding method, by means of which the features of a plurality of base stations are measured and the position of the object is located from these features, using a reference map, the apparatus comprising: at least one antenna for receiving the fields of the base stations and/or for emitting fields for receiving by the base stations as well as an evaluation unit for evaluating the measured features of the base stations. The evaluation unit is configured such that the abovedescribed method according to the invention can be carried out by means of it. Exemplary embodiments of the invention will be described in detail below with reference to the attached drawings, in which: Fig. 1 shows a schematic representation of the propagation models which can be used according to the invention; Fig. 2 shows a representation of a first building environment in which the method according to the invention was tested; Figs. 3 to 5 show diagrams which show the results of the tests of the inventive method in the first building environment; Fig. 6 shows a second building environment in which the method according to the invention was tested; and Figs. 7 to 9 show diagrams which show the results of the tests of the inventive method in the second building environment. In the embodiments of the inventive method described below, a mobile object is considered which has at least one antenna, by means of which the field strengths of a plurality of base stations in the area surrounding the object can be received and measured. Here, each base station transmits wireless signals with a certain power (i.e. with a certain fieldstrength value). These wirelessly transmitted signals propagate in space, where they are reflected, bent and absorbed, notably by persons and other objects in the space, such as e.g. walls, machinery or office furniture, in the surrounding area. In the exemplary embodiments of the inventive method described here, the propagation profile for each base station is described in the form of a twoor threedimensional map, each position in the map being uniquely linked to a received fieldstrength value for each base station. The received fieldstrength value is also referred to below as the RSS value (RSS = received signal strength). Consequently, such a map links position coordinates to RSS values of base stations, and the use of such maps, which are also referred to as "radio maps" is known from the prior art. Such maps are referred to below as location fieldstrength maps (RM for short) and they correspond in the special variant of the inventive method described here to the reference maps defined in the claims. The map describes a surface in a continuous space and can either be a continuous representation of e.g. a set of functions or a discrete representation as a matrix of recorded values which were taken from the continuous map. A location fieldstrength map is used below which constitutes a discrete representation of fieldstrength values at different positions or support points in the space. The location fieldstrength map RM is defined for K support points, a set of all the support points POSRM in the space being defined by the following matrix: where each individual term *POSRM = [kx™ *KRM] is the kth twodimensional position of this set at support points. A surrounding area comprising N base stations BS is considered below, it being possible for the location fieldstrength map RM to be represented as a set of K vectors of RSS values and for each of these vectors to have the dimension N for the field strengths of the N base stations. In this way, a matrix representation of the location fieldstrength map RM is produced, which reads as follows: RM = VRM PRMI PRM2 A PBMN VRM PSMI 2 A PRMN M M M O M //'RM. . PRMI AIM 2 A K PRMN_ (2) Here, the expression *PRM is the kth vector of RSS values from the set of K support points and the expression kpmn is the fieldstrength value for the base station n of the vector *PRM. Where there is a fixed number of support points, the support points must still be distributed suitably in the space. There is the possibility here of using a grid with an even spacing, the distance between the grid points being known. The positions of previously measured training data can also be used as the grid points. In order with the aid of the abovedefined location fieldstrength map now to locate the position from measured fieldstrength values at the object, a fingerprinting method (also referred to as "pattern matching") is used in the embodiment described here. This method is adequately known from the prior art and is therefore only outlined briefly below. The location fieldstrength map describes the field strength as a function of the position, i.e. it maps the two or threedimensional position vectors into Ndimensional fieldstrength vectors. The converse problem, i.e. determining a position from the fieldstrength vector, is referred to as localization estimation or localization and constitutes the positionfinding step as, for example, defined generally in step b) of claim 1. Typically, there is no analytical description of the inverse function by means of which the location field strength map was generated. The localization task can therefore be defined as follows: Starting from a location fieldstrength map RM, which describes the fieldstrength distribution in a space for N base stations BS and for a predetermined measured fieldStrength Vector pteas = [Pmeasl Pmeas2 ... PmeasAriT, which was determined at an unknown position posmea8 = [xmeas Ymeas] i the position *posRM in the location fieldstrength map RM must be found so that a predetermined cost function between pbeaa and the fieldstrength vector kpRM is minimized. This can be achieved in various ways, but patternmatching algorithms are most suitable for the task set here, as Pmeas can be viewed as a pattern that has to be mapped onto a given model, represented by the location fieldstrength map RM. In the embodiments described here, the patternmatching algorithm uses a sequential pass through the search space in order to search for the particular position *posRM in the location fieldstrength map that minimizes the following cost function. Here, the index h stands for the hth position ^posR*, in the search space and n is the index of the base stations. Pmeas n is the nth element of the measured fieldstrength vector Pmeas/ and 'PRM n is the nth element of the fieldstrength vector hpstt, which originates from the location fieldstrength map RM. The index H at which the cost function r is minimal, i.e. H = argmin(r), returns the nearest spatial position in the model (i.e. in the location fieldstrength map) for this pattern. The expression "nearest" relates directly to a distance measurement which is defined in the cost function. The equation (3) returns, for example, the square of the Euclidian distance in Ndimensional space. The value "pos^ as the nearest match with the search pattern marks the "nearestneighbor algorithm" (NN) just described. The above method is used in all the embodiments described below in order to determine the located position from the measured fieldstrength vector The method according to the invention is distinguished in that initially a location fieldstrength map determined with simple means can be employed for positionfinding, the fieldstrength map being updated correspondingly during at least some of the positionfinding steps in order to adapt it precisely to the prevailing surrounding area. The way in which the location fieldstrength map that exists initially can be determined in the embodiments described here will be described below. When initially determining the location fieldstrength map only the reciprocal measurements for each base station are used. With these few measurements, the initial model is not very accurate, but the model can be generated rapidly without a technician having to be deployed to measure the location fieldstrength map. The generation of the initial model in the embodiments described here consequently represents a kind of selfcalibration of the localization system. The selfcalibration is independent of the steps described further below for updating the location fieldstrength map. To generate the initial location fieldstrength map, theoretical fieldstrength propagation models are used which usually define the fieldstrength value as a function of the distance and are characterized by the following equation: Here, p represents the power (the field strength) in a logarithmic representation, the field strength depending on the distance d and the fixed parameters of the output power po and the loss factor y. For the function f (d) , 10 log (d) is usually used, as defined in the printed publication [1], a minimum distance d0 having to be defined for use of the model in order to avoid singularities. The minimum distance presents a problem with wave propagation in buildings since there a receiver can be very near to a base station. Furthermore, the initial model for wave propagation in buildings constitutes a very marked approximation, so that even a simpler wavepropagation law than according to equation (4) does not generate greater errors, but leads more easily to an estimate from the few measurement values. A linear function f (d.) =d has therefore been chosen in the embodiments below. Based upon equation (4) above, two different models are used in the embodiments of the invention described here in order to establish the fieldstrength vectors at a plurality of support points in the location fieldstrength map during initialization of the method. In particular, the radial model and the DPM model (DPM = dominant path model) are considered, both models being adequately known from the prior art. In the radial model, a radial symmetry is assumed, so that equation (4) is applied directly for each base station BS, which in turn means that the field strength received from a mobile object depends exclusively on the distance to the respective base station. During initialization, each base station BS measures the RSS value of the Nl neighboring base stations such that the two parameters po and 7 for each base station can be determined by means of the least squares method ("least squares fit") . The DPM model described below, which is known for example from printed publications [4] and [5] , is based on an improvement of the MWM model (MWM = multi wall model) known from printed publication [1] . The MWM model uses a modified variant of equation (4), which also takes into account the walls through which the fieldstrength signal has to pass on the direct path between a transmitting base station and certain measuring points. The modified equation reads as follows: Here, the expression WAF represents the wall attenuation factor, which is a parameter that depends on the thickness of the wall and the dielectric properties of the wall material. In general, each wall can have its own WAF factor. In a simple model, different wall types can be defined. By means of the summation defined in equation (5) , propagation paths in which more than one wall is traversed can also be taken into account. In order to use the model characterized by equation (5), a description of the position and orientation of the walls and/or objects in the space must be available, for example in the form of a building plan which exists as an image in known graphic formats (e.g. jpeg, gif, etc.) or is made available by a CAD system. The greatest disadvantage of the MWM method is that it considers only direct paths between receiver and transmitter. A direct path does not necessarily have to match the particular path on which the strongest field strength is received (socalled dominant path). Consequently, the result is often an estimation of an RSS value that is lower than the actual RSS value at the object position under consideration. In order to resolve this problem, raytracing algorithms, adequately known from the prior art, compute all the possible paths between transmitter and receiver and summate the individual contributions of these paths, by which means a better estimation of the field strength is obtained than by the MWM method alone. However, the computation of all the paths requires a very large amount of computing time for each position. In particular, not only is the number of paths large but a high level of accuracy of the map under consideration is needed for the algorithm. The DPM model lies between the two techniques of the MWM method and the raytracing algorithms. It uses information from the map of the surroundings to determine the dominant path that makes the greatest contribution to signal strength in the overall Signal. The DPM path can be longer than the MWM path and can also continue through walls. In the DPM method, only the dominant path is used in order to estimate the received field strength in an object position under consideration. The model is, furthermore, described by equation (5), but only the length of the dominant path and the corresponding walls on this path are considered. Better results are obtained by means of the DPM method than with the MWM method, and it has been shown that the results are almost as good as the results which are determined with the timeconsuming raytracing algorithms. Taking a simple building plan as an example, Fig. 1 shows how the methods described above differ from one another. Here, a corridor C and three rooms Rl, R2 and R3 which are separated by corresponding walls are shown. Furthermore, a base station BS is represented as a transmitter and an object point Rx as a receiver. The dotdash line shows the path which is used in the MWM method. This path goes through walls and is a direct path between BS and Rx. The dashed lines show the various possible signal propagation paths in the raytracing method. In the DPM method, the particular signal propagation path via which the greatest fieldstrength value arrives at point Rx is used for computing the field strength. This is the dominant path, which is reproduced in Fig. 1 as a solid line. It will be noticed that in the example shown in Fig. 1 this path crosses through just a single wall, namely the wall between the corridor C and the room R3. In the DPM model, the dominant path can be computed with the aid of the following steps (see also printed publication [6]): Step 1: Generate a vector structure map. Define the room structure and the room index. Generate a symmetrical room connection matrix Cl . If the room i and the room j share the same walls, set Cl(i, j) = 1, otherwise set Step 2: Detect the rooms in which the transmitter and the receiver are positioned, using the algorithm defined in the printed publication [7] . If the transmitter and the receiver are positioned in the same room, go directly to step 3 below. Otherwise search for the shortest room connection from the room with the transmitter to the room with the receiver in the matrix Cl with the aid of the Dijkstra algorithm known from the prior art (see printed publication [8] ) . The cost value for the shortest path is the number of walls penetrated. Step 3: Combine the room with the transmitter, the room with the receiver and the rooms in the shortest path by omitting the walls which the rooms share. Step 4 : Detect convex corners in the room combined in this manner and generate a symmetrical connection matrix C2 between the transmitter, the receiver and the convex corners. If there is no object between the point i and the point j, C2(i, j) corresponds to the distance between the point i and the point j. Otherwise, C2(i, j) is set to . Step 5: Find the shortest path between the transmitter and the receiver in the matrix C2 via the Dijkstra algorithm mentioned above. The cost value of the shortest path is given by the sum of the distances. This shortest path is called a curved path. Step 6: Use reciprocal fieldstrength measurements in order to estimate the parameters p0, Y and WAF in the DPM model with the aid of a leastsquare fit. step 7: Calculate the loss on the direct path between transmitter and receiver and for the curved path which was obtained in step 5. If the loss on the curved path is lower than on the direct path, the curved path is used as the dominant path. Otherwise, the direct path is used as the dominant path. Finally, the initial location fieldstrength map is obtained with the aid of the information from the geographical map and the DPM model. This is done by determining for each position in the location fieldstrength map the length of the dominant path, and from this the corresponding field strength at the respective position is then obtained by means of the formula (5) . A discrete location fieldstrength map is then available in which for each support point the locational position and the corresponding fieldstrength vector are indicated. The embodiments of the inventive method described here use a location fieldstrength map which has been generated in an arbitrary manner, whereby, in particular, one of the methods described above is used for generating such a location fieldstrength map. Here, the initial location fieldstrength map is only very rough and does not have the desired accuracy for localizing objects precisely. Therefore, according to the invention, an updating of the fieldstrength map is always carried out, e.g. just when the position of a mobile object is being located. In the embodiments described below, an online learning method is used in which new positionfinding processes are used in order to modify the initial location fieldstrength map and thereby improve localization accuracy. The learning method is a method which is based on Kohonen' s self organization map (see printed publication [2]) . The updating is achieved by means of a correction term for the error, this correction term being called an "update surface". This update surface is added to the individual fieldstrength values of the location fieldstrength map, in order by this means to improve the accuracy of the fieldstrength values at the individual support points . The update surfaces are determined for each base station BSn and are based on the local error Apn, which is defined as: = Pnwaa n ~ Pstt nr (6) where praeas n is the measured field strength and pm „ the field strength which is predetermined by the location fieldstrength map, specifically e.g. the field strength at the particular map position which was determined by position finding using the fingerprinting method. The center of the update surface is defined by a specific position in the location fieldstrength map, which is designated °poa = [°x °y]. The determination of °pos depends on the way in which the localization is carried out, as described in detail further below. The maximum value of these update surfaces is scaled by a factor K, which is considered to be the learning rate, and the absolute value of the update surface decreases evenly in all directions and ultimately reaches zero. Consequently, the area of influence of the update surface is spatially limited by a surrounding area, which is referred to below as the adaption width or adw. The update surface is generally defined as: In accordance with this definition, the update surface combines two important information sources, namely on the one hand the local error Ap of the fieldstrength value and on the other the weighting factor /(/•) that reflects the relevance of the error as a function of the distance. The distance or the radius r between the central position °pos and any function pos = [x y] is defined here as: (Equation Removed) The weighting factor is defined in the embodiment described here by a Gaussian surface as follows: This weighting factor has already been described in document [2], Since a Gaussian bell is at three times the standard deviation, i.e. at 3a, essentially zero from the center of the bell, the function as per equation (9) is, in a preferred embodiment of the invention described here, approximated by the following function: According to this function, an adaption width adw is established which preferably lies at 3a. All function values outside this adaption width are set to zero. This means that fieldstrength values for positions having a distance greater than 3a from the center of the Gaussian bell are not updated. Use is made here of the fact that a local error is of relevance only in a certain area surrounding the measuring point under consideration since, in the case of support points further away, additional effects can falsify the measurement. During the learning method, in particular during gen eration of the update surface, there are two options for determining the local error, which is defined ac cording to equation (6). According to one variant, the central error is considered at which the measured field strength Pbaa» = °P is linked to the located position posmC'p) in the fieldstrength map. Here, the located position is preferably determined by means of the fingerprinting method. This located position is used as the center of the update surface, i.e. °pos = posVM(°p) . The local error *Ap is calculated at each position *POSRM in the location fieldstrength map for all base stations as follows: where the measured fieldstrength vector °p defines the center °pos. "PRM is the fieldstrength vector at the located position in the location fieldstrength map determined in particular via the fingerprinting method. The index k = {1,2,...,K} represents all positions *POSRM which exist in the location fieldstrength map. Where the central error is used, *Ap = Ap is constant for all k and depends only on the choice of °pos. The set of all *Ap for all K support points in the location fieldstrength map then generates update surfaces having constant values, more precisely one update surface for each base station respectively. The localization error Ap consequently constitutes a vector with constant multipliers in equation (7). By means of a small change to equation (11), the central error is converted into a neighborhood error, which is defined as follows: (Equation Removed) where *Ap is now no longer constant in the space since °p is compared with a set of K fieldstrength values which originate from the location fieldstrength map. For reasons of efficiency, only the positions which lie in the region that is limited by the above adaption width adw are used here for calculating both the central and the neighborhood error. In accordance with the method just described, an updating of the location fieldstrength map can be achieved by Gaussian functions whose center is given by a located position which was determined with the aid of the fingerprinting method described above. This is unsupervised learning, as the actual position of the object is not known at any time, rather the located position is always used as the center of the update surface. In a further variant of the invention, the above method can, however, also be used for supervised learning, as will be described below. A pathpresetting method (also called PPS method) will be examined below. This method serves in particular in the calibration of a location fieldstrength map. A technician passes with the mobile object through the area of the location fieldstrength map, it being known during passage through the area at what positions the mobile object is located. The technician passes from one base station to the next and, in doing so, has to indicate which base station he starts from and which base station he arrives at. All positionfinding measurements between the base stations take place on a defined path and at defined measuring points, a straight line with evenly spaced measuring points being considered for test purposes. In contrast to unsupervised learning, for each newly measured fieldstrength vector °p the position °pos is known directly as a position on the defined path. This known position is deployed as the center of the update surface used. Since the position °pos is based on information known in advance, this calibration method is called supervised learning. The choice of paths which are executed by the mobile object during calibration is designed to take account of the requirement that as large an area of the location fieldstrength map as possible is covered. In particular, paths which extend over long distances return a better learnt dataset than paths which connect only closely neighboring base stations. The sequence of the paths executed therefore has a direct effect on the quality of the calibration. For N base stations BS, (N2N)/2 different direct connecting paths exist, and a preferred variant will be described below specifying in which order the mobile object should execute the paths in order to achieve good calibration. First, the matrix path(ifj) will be defined, which comprises Q ordered collinear positions between the base stations BSi and BSj (posBsi = ^pos, posBSj = Qpos) . Each position qpos in the path path^,^ (where q = {1,2,...,Q}) also defines the position of the center of the update surfaces used, i.e. °pos = cqpos. The path path(i/^) can be written as a vector as follows:path(lj] = 1 pos ' clpos ' 2 cl pos pos M = M Q'pos **pos Q pos ^pos Here, each entry represents a two or threedimensional position of the measuring points on the path. As previously mentioned, the individual positions are evenly spaced so that the distance Apath(i,j) between consecu tive positions in the lows :can be calculated as fol (141 In order to achieve good coverage of the location fieldstrength map by paths executed, a socalled pathcost function is examined, in which firstly the distance from each position qpos assumed on a path to the next base station is examined. This distance is designated qdBs and reads as follows: (Equation Removed) With the aid of this function, a cost value rpath for a path is computed whereby the maximum is selected from all the distances as per equation (15) for a path between the base stations i and j . This gives rise to the following cost function: (Equation Removed) This pathcost function reflects how far removed the area through which the path runs is from base stations. In order to achieve good calibration, according to the variant of the inventive method described here, paths with high path costs are run through at first. In other words, the paths are run through by the mobile object in decreasing order of their cost. It should be noted here that the selection of the paths can itself be affected by certain restrictions, for example by the fact that positions which have already been assumed previously are dynamically taken into account in the computation of new path costs. Each path can now be viewed as a set of measuring points, each of these measuring points being linked to a local error as per equation (6) described above. For this reason, in the embodiment described here, a path is viewed as a whole and adapted blockwise, whereas in the embodiment described previously an adaptation of the location fieldstrength map can be carried out in a measuring point after each positionfinding process. In the embodiment being described now, however, all the positions which are associated with a path under consideration are taken into account when computing the update surface. Consequently, the update always takes place only after the traversing of a path by the mobile object. In the variant described here, the Gaussian area is again used as weighting of the local error as per equation (9). This area is based on the central error which is given by equation (11) . In order to take into account simultaneously all the positions on a path between the base stations BSi and BSj, some modifications have been made in the composition of the update surface . The first modification is that the standard deviation a has been made proportional to the even spacing between the positions on the path, i.e. proportional to Apath(i,j). Consequently, a = for a fixed $, provided that Apath(i/j) is different for each path. The sequence of the positions along the path generates a sequence of Gaussian bells for each position on the path, the Gaussian bells overlapping one another. A type of normalization should therefore be used in order to achieve the outcome whereby the sum of all the Gaussian bells along the path lies within the limits of the original local error values on the update surface. The normalization used here is based on socalled "partitioningtoonenormalization", which is used in radial networks (see printed publication [3]). By analogy with equation (8), the radius r for the central position °poa and for the qth position cqpos on a path is defined as gr as follows: Accordingly, the local error defined in equation (11) is modified for a sequence of path positions c Using gr and ?Ap in equation (7) or (9) yields in the following term as normalized update surface for the qth path position qpos and for the base station n: Here, £ is a positive term so as to ensure that the denominator never becomes zero, which means that the normalized Gaussian surfaces also actually decrease to zero at positions at a large distance from the center. It must be taken into account here in particular that the choice of £ also introduces deformations in the surface. If £ is too large, the surface rescales to smaller values. If £ is too small, the surface assumes rectangular shapes instead of a bell, even though the surface is normalized to one. It has been shown that a preferred value for £ lies at 0.3. The essential difference between the calibration just described using the PPS method and calibration using unsupervised learning is that, in the case of unsupervised learning, for each new measurement of the fieldstrength value °p the position °pos is derived exclusively from the model used in positionfinding, in particular by means of the fingerprinting method. This means that the positions estimated by means of the originally used location fieldstrength map are used in order to update the model (i.e. the location fieldstrength map) , and this is done without the actual position of the object being known. A further important difference between unsupervised and supervised learning is that, in the case of unsupervised learning, the measurements can be undertaken at any location in the location fieldstrength map and are not restricted only to direct paths between two base stations, as is the case with the PPS method. Consequently, with learning more points can be taken into account. A further difference is that, in the case of unsupervised learning, updating of the location fieldstrength map is carried out with each new measurement, whereas updating in the PPS method cannot be carried out until a path has been executed. Experimental results are presented below which are based on experiments conducted by the inventors, by means of which experiments objects were localized using previously described variants of the inventive method. The method according to the invention was tested for different surroundings in office buildings. As an initial model, both the radial model and the DPM model were used for each surrounding area and both for the supervised learning method and for the unsupervised learning method. To determine the localization error, the entire measurement was examined. The mean error in the position estimation was calculated. The error itself is the distance between the estimated position and the actual position at which the measurement was carried out. Each position was measured three times and the average formed from these measurements. For the supervised learning, a location fieldstrength map based on a regularly spaced grid was used, each support point being offset from the other by one meter. The actual measuring positions were not exactly collinear with the direct paths between the base stations. However, the positions were chosen such that they lay as close as possible to the direct path. It could thus be assumed that Apath(i,j) was constant for each path. The localization error was always calculated when a single path between two base stations was executed by the mobile object. In the diagrams explained further below, the number of the path is consequently reproduced on the X axis. For the unsupervised learning, the support points of the location fieldstrength map were formed by the actual measuring positions. The localization error was calculated with each new measurement. The measurements were chosen randomly from a set of measurements, repeated measurements being permitted without restriction. The control parameters K and $ were varied both for the supervised and for the unsupervised learning. Fig. 2 shows the plan of a first building OF1 in which the method according to the invention was carried out. Nine base stations BS1, BS2, ..., BS9, which are indicated by stars, are installed in the building. Measurements were carried out in the building at a total of 223 different measuring points, these measuring points being indicated in Fig. 1 as crosses. The base stations are DECT standardcompliant transceivers. For the PPS method, a combination of all the possible paths between the base stations yields 36 different paths. If only the measuring points along the paths are considered, the measurement consists only of 88 different measuring points. The order of the paths executed was found via equation (16) above. For the first 10 paths with numbers 1 to 10, the following base stations were connected to one another: Path no. 1: from base station BS8 to base station BS3 Path no. 2: from base station BS3 to base station BS1 Path no. 3: from base station BS1 to base station BS8 Path no. 4: from base station BS8 to base station BS2 Path no. 5: from base station BS2 to base station BS9 Path no. 6: from base station BS9 to base station BS8 Path no. 7: from base station BS8 to base station BS6 Path no. 8: from base station BS6 to base station BS3 Path no. 9: from base station BS3 to base station BS2 Path no. 10: from base station BS2 to base station BS1 Fig. 3 shows diagrams which reproduce the localization error LE and the parameters K and $ for the PPS method depending on the number of the path PN for the building OF1. In accordance with the algorithm described in the printed publication [2], the control parameters K and $ were chosen such that they diminish over time. According to Fig. 3, the localization error was compared both for the radial propagation model RAD and for the DPM model DPM as initial models. It will be recognized that the method returns smaller localization errors LE if the initial location fieldstrength map was computed by means of the DPM model. For both models, the localization error LE became increasingly small as the number of paths executed increased. Fig. 4 shows a representation analogous to that of Fig. 3, in which the unsupervised learning method was used for learning for the building OFl. Here, all 223 measuring positions were used for learning and the learning process was carried out for 5,000 randomly selected measuring points. In contrast to Fig. 3, the individual iterations I, i.e. the individual measurements at the measuring points, are now reproduced along the X axis. Again, cases were compared in which, as original location fieldstrength map, a map according to the radial model RAD or a map according to the DPM model DPM was computed. Fig. 5 shows the same results as Fig. 4, but the iterations are reproduced on a logarithmic scale, so that the trend for the first 1000 iterations can be better seen. It will be recognized that the localization error decreases as the number of measurements increases. The learning is particularly fast at the start when the location fieldstrength map generated by means of the model is not yet so good. Toward the end, the learning becomes slower and the accuracy reaches the maximum possible accuracy. Maximum accuracy is given by the accuracy of the fingerprinting method used, by means of which the position is measured. In the fingerprinting method, the points at which the algorithm was learned unsupervised are used as reference points with known positions. The localization error of the fingerprinting method is limited at the bottom end, namely by the measurement noise of the fieldstrength measurements (due to signal shadowing, the noise of the measuring device and other environmental factors) and due to the density of reference points and the number of base stations and other effects. The maximum possible accuracy for the building OF1 lies at 4.6 meters. Fig. 6 shows the plan of a second building OF2, for which the method according to the invention was likewise tested. 14 WLAN base stations BS1, BS2, ..., BS14 are installed in this building. The base stations are again reproduced as stars. The complete set of measurements contains 114 different measuring points, which are represented as crosses. Again, the supervised PPS learning method was implemented. For this method, a combination of all possible paths between all the base stations yields a total of 91 paths. The order of the paths executed was again computed with the aid of equation (16) above, the concave shape of the building also being taken into account. This means that paths which pass beyond the limits of the building were not taken into account, for example the path from base station BS3 to base station BS10. For the first ten path numbers 1 to 10, the following connections from base stations were used: Path no. 1: from base station BS2 to base station BS9 Path no. 2: from base station BS9 to base station BS12 Path no. 3: from base station BS12 to base station BS8 Path no. 4: from base station BS8 to base station BS7 Path no. 5: from base station BS7 to base station BS14 Path no. 6: from base station BS14 to base station BS12 Path no. 7: from base station BS12 to base station BS13 Path no. 8: from base station BS13 to base station BS5 Path no. 9: from base station BS5 to base station BS11 Path no. 10: from base station BS11 to base station BS3. Fig. 7 shows again the localization error LE and the model parameters K and sitions from the original set of measurements. Analogously to Fig. 3, the control parameters K and $ decrease over time. Fig. 8 shows the result for unsupervised learning, all 114 measuring positions being included. The learning method was implemented using 5000 randomly selected measuring points. Fig. 8 shows the results on a linear scale of the X axis and Fig. 9 shows the same results on a logarithmic Scale of the X axis. It will be recognized that the use of an original location fieldstrength map according to the DPM model returns significantly better results, as in this model additional information relating to the layout of the surrounding area, in particular the positions of the walls, is taken into account. The building environment OF2 is characterized here in particular in that the walls are coated with metal foil, which greatly attenuates the signals. Accordingly, the information about the walls provides a very good starting model, which results in a low localization error even before the start of the learning method. The learning consequently beings no substantial improvement in the original localization error. By comparison, the method using the radial model starts with an initially very poor localization error, which then, however, improves significantly during learning. As is evident from the preceding comments, a very good level of localization accuracy is achieved by means of the localization method according to the invention, the installation and maintenance costs for the method being very low. In particular, no initial calibration of the method with the aid of reference measurements is re quired, as the calibration is carried out by updating the location fieldstrength map during the method itself. The method starts here with an initial wave propagation model of low complexity, the parameters of which have been determined through a small number of measurements, for example through selfmeasurement by the base stations. It will be recognized that, where the DPM model is used, the accuracy of the method is better than where the simple radial model is used. The original location fieldstrength maps are according to the invention adapted online during the learning method, a distinction being made between an unsupervised learning method and a supervised learning method. In the unsupervised learning method, the localization and the updating of the location fieldstrength map are carried out simultaneously. In the supervised learning method, by contrast, a known path is predetermined on which the mobile object moves. The accuracy achieved by means of the method is comparable with localization algorithms which use reference measurements; however, it is not necessary in the case of the method according to the invention to carry out such reference measurements before the start of the method. Bibliography: [1] T. S. Rappaport, Wireless Communications Principles and Practice, 2nd Ed., Prentice Hall Inc., 2002. [2] T. Kohonen, "The selforganizing map," Proceedings of the IEEE, vol. 78, no. 9, pp. 14641480, September 1990. [3] M. Schlang, M. Haft and K. A. Fuchs, "A comparison of RBF and MLP networks for reconstruction of focal events from bioelectric/biomagnetic field patterns," in Mustererkennung 1994 Erkennen and Lernen, 16th DAGM Symposium and 18th Workshop of the OAGM, 1994. [4] G. Wolfle, R. Wahl, P. Wertz, P. Wildbolz and F. Landstorfer, "Dominant path prediction model for indoor scenarios", German Microwave Conference (GeMIC) 2005, April 2005. [5] G. Wolfle, and F. Landstorfer, "Dominant paths for the field strength prediction," 48th IEEE Vehicular Technology Conference (VTC) 1998, pp. 552556, May 1998. [6] H. Wang, Fusion of Information Sources for Indoor Positioning with Field Strength Measurements, Masters Thesis, TUMunich, 2005. [7] M. Shimrat, "Algorithm 112: position of point relative to polygon," Communications of the ACM, vol. 5, no. 8, pp. 434, August 1962. [8] Robert W. Floyd, "Algorithm 97: shortest path," Communications of the ACM, vol. 5, no. 6, pp. 345, 1962. We Claim: 1. Method for the computeraided localization of a mobile object with the aid of a featurebased positionfinding method, by means of which the features of a plurality of base stations (BS1, ...,BS14) are measured, and the position of the object is located from these features, using a reference map (RM), the feature of a base station (BS1, ..., BS14) characterizing a field emitted by the base station (BS1, ..., BS14) at the object or a field emitted by the object at the base station (BS1, ..., BS14), wherein: a) during initialization of the method, a reference map (RM) comprising a multiplicity of support points, which are each represented by a position (*POSRM) and a featuredependent value (*PRM) assigned to the position (*POSRM) / is predetermined; b) during use of the method a plurality of position finding processes are carried out, by means of each of which a measured featuredependent value (°p) and, from this, a located position (posRM(cp)), of the object are determined via the featurebased positionfinding method, using the predetermined reference map (RM); c) one or more updates of the predetermined reference map (RM) are carried out in each case for at least some of the locations, during which updates the featuredependent values (*PRM) are each corrected by a correction term (SUPD) at the support points of the reference map (RM) in a predetermined area surrounding an object position (POSRM(°P)), the correction term (SUPD) depending on the difference between the measured featuredependent value (°p) for the respective location and the feature dependent value (*PRM) at a support point in the predetermined surrounding area. 2. The method as claimed in claim 1, wherein the object position (POSRM(°P)) is previously known or is the posi tion located by means of the respective position finding process. 3. The method as claimed in claim 1 or 2, wherein the step c) is carried out at the time of the respective positionfinding process or after a predetermined num ber of positionfinding processes has been carried out. 4. The method as claimed in any one of the preceding claims, wherein in step a) the predetermined reference map (RM) is computed by means of a wave propagation model. 5. The method as claimed in claim 4, wherein the wave propagation model is a radial propagation model. 6. The method as claimed in claim 4, wherein the wave propagation model is an MWM propagation model (MWM = multi wall model) or a DPM propagation model (DPM = dominant path model). 7. The method as claimed in any one of claims 4 to 6, wherein the parameters of the propagation model are de termined at least partially by means of a self measurement by the base stations (BS1, ..., BS14). 8. The method as claimed in any one of the preceding claims, wherein each of the plurality of position finding processes in step b) is carried out with the aid of a fingerprinting method and/or with the aid of a probabilistic method. 9. The method as claimed in any one of the preceding claims, wherein the size of the correction term (Sapo) decreases in magnitude with increasing distance from the object position (posKt(°p) ) . 10. The method as claimed in claim 9, wherein the cor rection term (SOPD) depends on a Gaussian function (f(r)) with its center in the object position 11. The method as claimed in claim 10, wherein the pre determined area surrounding the object position (POSRM(°P) ) depends on the standard deviation (a) of the Gaussian function (f(r)). 12. The method as claimed in claim 11, wherein the pre determined surrounding area is formed by a circle about the center of the Gaussian function (f(r)) having a ra dius which is greater than or equal to the standard de viation (a) and is, in particular, a multiple of, pref erably three times, the standard deviation (a) . 13. The method as claimed in any one of claims 10 to 12, wherein the standard deviation (a) of the Gaussian function (f(r)) depends on the spacing of the support points on the reference map (RM) and lies in particular between the supportpoint spacing and five times the supportpoint spacing. 14. The method as claimed in any one of claims 10 to 13, wherein the Gaussian function (f(r)) is approxi mated by a function which decreases to zero with increasing distance from the center and, in particular, is zero at a value greater than or equal to the standard deviation (a) , preferably at three times the standard deviation (a) . 15. The method as claimed in any one of the preceding claims, wherein the measured featuredependent value (°p) is a feature vector, in particular a fieldstrength vector, measured at the object or at the base stations at the time of the positionfinding process and wherein the predetermined reference map (RM) is a location fea ture map in which the featuredependent values are fea ture vectors (*PRM) for the positions (POSRM) of the support points, a feature vector comprising as entries the features of the base stations (BS1, ..., BS9) for the corresponding position. 16. The method as claimed in claim 15 in combination with any one of claims 9 to 14, wherein the correction term SOPD for a support point k inside the predetermined surrounding area is given by the following function or depends on the following function: where f (r) is a Gaussian function defined as follows: where a is the standard deviation and r the distance from the object position (poswi(cp) ) and K is preferably a value less than 1 and particularly preferably less than or equal to 0.5, where kAp is a differential feature vector comprising one entry in each case for the feature of a base station from N base stations, where the differential feature vector *Ap is either defined as: where °p is the measured feature vector at the time of the positionfinding process and "PRM is a feature vector determined at the located position by means of the location feature map (RM) , or is defined as: *Ap = °p  where °p is the measured feature vector at the time of the positionfinding process and *PRM is the feature vector at the support point k of the location fieldstrength map. 17. The method as claimed in claim 16, wherein the path of the mobile object is predetermined and the object position (POSRM(CP) ) at the time of a respective posi tionfinding process is a previously known position. 18. The method as claimed in claim 17, wherein the ob ject moves on one or more predetermined path sections between, in each case, two base stations (BSl, ..., BS14), object positions on the respective path section being known in advance and a positionfinding process being carried out at each of these object positions. 19. The method as claimed in claim 18, wherein, at object position q (q = 1, ...,Q) of the Q object positions on a path section, the correction term SUED (r, q) is defined as follows: where a is the standard deviation and qr the distance from the object position q and gr the distance from the object position g and where K is preferably a value less than or equal to 0.5, where ^ is a positive term which ensures that the denominator does not become zero, where I, is preferably less than 1 and is, in particular, 0.3, where 'Ap is a differential feature vector comprising one entry in each case for the feature of one base station of N base stations, where the differential feature vector 'Ap is defined as: where oqp is the measured feature vector at the time of the positionfinding process at the object position q and C 20. The method as claimed in claim 19, wherein the dis tance (Apath(i,j)) between consecutive object positions on a predetermined path section is constant and the standard deviation (a) depends on the constant distance (Apath(i/j)) and is, in particular, directly proportional to the constant distance (Apath(i, 21. The method as claimed in any one of claims 18 to 20, wherein the object executes a sequence of predeter mined path sections, the order of the path sections executed being determined with the aid of a cost func tion for each path section. 22. The method as claimed in claim 21, wherein the cost function for a respective path section indicates how great the maximum distance of an object position on the path section is from the nearest base station (BS1, ..., BS14) or from an already traversed nearest object posi tion. 23. The method as claimed in any one of the preceding claims, wherein the featurebased positionfinding method comprises a fieldstrengthbased position finding method and/or a runtimebased positionfinding method and/or anglebased positionfinding method. 24. The method as claimed in claim 23, wherein, in the case of the runtimebased positionfinding method, only such fields are taken into account as exhibit a minimum signal strength at the object. 25. The method as claimed in any one of the preceding claims, wherein the featurebased positionfinding method uses the fields of a DECT and/or WLAN and/or mo bile radio network for positionfinding. 26. An apparatus for the computeraided position finding of a mobile object with the aid of a feature based positionfinding method, by means of which the features of a plurality of base stations (BS1, ...,BS14) are measured and the position of the object is located from these features, using a reference map (RM), the feature of a base station (BS1, ..., BS14) characterizing a field emitted by the base station (BS1, ..., BS14) at the object or a field emitted by the object at the base station (BS1, ..., BS14) , comprising at least one antenna for receiving the fields of the base stations (BS1, ..., BS14) and/or for emitting fields for receiving by the base stations and an evaluation unit for evaluating the measured features of the base stations (BS1, ..., BS14), the evaluation unit being configured such that a method as claimed in any one of the preceding claims can be carried out by means of the evaluation unit. 27. A method for the computeraided localization of a mobile object with the aid of a featurebased position finding method substantially as hereinbefore described with reference to the accompanying drawings. 28. An apparatus for the computeraided position finding of a mobile object with the aid of a feature based positionfinding method substantially as herein before described with reference to the accompanying drawings. 

1708delnp20081Abstract(31102013).pdf
1708delnp20081Claims(31102013).pdf
1708delnp20081Correspondence Others(31102013).pdf
1708delnp20081Drawings(31102013).pdf
1708delnp20081Form3(31102013).pdf
1708DELNP2008Assignment(28102014).pdf
1708DELNP2008Claims(28102014).pdf
1708delnp2008Correspondence Others(16072012).pdf
1708delnp2008Correspondence Others(31102013).pdf
1708DELNP2008CorrespondenceOthers(28102014).pdf
1708delnp2008CorrespondenceOthers(29092014).pdf
1708delnp2008correspondenceothers.pdf
1708delnp2008description (complete).pdf
1708delnp2008GPA(16072012).pdf
1708DELNP2008Petition137(28102014).pdf
1708delnp2008Petition137(31102013).pdf
Patent Number  264055  

Indian Patent Application Number  1708/DELNP/2008  
PG Journal Number  49/2014  
Publication Date  05Dec2014  
Grant Date  02Dec2014  
Date of Filing  27Feb2008  
Name of Patentee  SIEMENS ENTERPRISE COMMUNICATIONS GMBH & CO. KG  
Applicant Address  HOFMANNSTR. 51, 81379 MUENCHEN, GERMANY  
Inventors:


PCT International Classification Number  H04Q 7/38  
PCT International Application Number  PCT/EP2006/069616  
PCT International Filing date  20061212  
PCT Conventions:
