Title of Invention

"METHOD FOR SECURING A DATA PACKET STREAM TRANSMITTED BY A TRANSMITTER DEVICE"

Abstract

Method for securing a data packet stream transmitted by a transmitter device (1) comprising computation means (2) and means for applying said computation means (2) to each row and each column of a matrix, comprising a step of: - arranging the packets to be sent in a packet matrix of D rows and L columns; characterized in that it further comprises following steps: generating a correction packet, for each row and each column of the matrix, by said computation means (2) by applying an error correction function to each row and each column of the matrix, said correction packet enabling the reconstruction of at least one lost packet in the row, respectively the column, to which the error correction function is applied; -sending the resulting correction packets in addition to the data packets. FIG. 1

Full Text

The present invention relates to a method for securing a data packet stream transmitted by a transmitter device. The present invention concerns error correction when sending digital data in the form of packets. When sending digital data over a network in the form of data packets, the packets may be affected by transmission errors. Networks enabling the transfer of digital data in the form of packets have particular characteristics, such as bandwidth, latency or reliability, that vary from one type of network to another. Depending on the type of network, different levels of sensitivity- can be observed for different types of errors likely to occur during data transmission by packet over this network. The types of errors likely to arise include the transmission of a packet with an erroneous bit, the random loss of a packet, the loss of a sequence of adjacent packets, the duplication of a packet or of a sequence of packets. Traditionally, there are two types of method for addressing these errors. First, there are the error detection methods for detecting that a transmitted packet is in error and for ignoring it. Generally these methods are used when the transmission protocol has the facility to request that the lost packet be resent. Examples of such methods include the CRC (Cyclic Redundancy Check) method and the TCP (Transmission Control Protocol) protocol on IP networks. Secondly, there are methods enabling not only the detection of an erroneous packet, but also its reconstruction. These methods are known under the acronym FEC (Forward Error Correction) and are well suited to protocols in which it is not possible to request that an erroneous data packet be resent. An example of the use of such methods occurs in the case of protocols for transmitting multimedia content in real time, where meeting the real-time requirement is incompatible with resending erroneous packets. The RTP (Real Time Protocol) protocol is an example of such a case. The general principle of operation of FEC methods is applying a function, for example an XOR, to a group of data packets. The result of this function produces a packet, called a correction packet, which is transmitted in addition to the data packets used to generate it. When a packet is detected as being in error, the correction packet, associated with the correctly transmitted packets, enables the erroneous packet to be reconstructed. However, this method has its limitations in that there are some errors that cannot be corrected. Specifically, if there are two erroneous packets in the group of packets used to generate a correction packet via the XOR function, that correction packet may not be sufficient to reconstruct the two erroneous packets. The strategy adopted in choosing packets for generating the correction packet is therefore crucial to the effectiveness of the correction method. In particular, this strategy will depend on the type of errors occurring most frequently on the type of network used. For example, computing the correction function on N adjoining packets will result in correcting random errors that are few in number, but will be ineffective on a series of lost packets. To address series of lost packets, a known approach, as documented for example in the Pro-MPEG forum's "Code of Practice", is to compute the correction function on a group of D packets taken periodically every L packets. The function is applied to packets i, i+L, i+2L i+(D-1)L In this way, a series of erroneous packets, of at most L adjoining packets, may be corrected by the method. The problem with this method is that, although very well suited to series of erroneous packets, it turns out to have a statistical rate of success that decreases very rapidly as the rate of random errors increases. The need is therefore apparent of a method that better withstands an increase in the rate of random errors while maintaining good correction performance for series of data packets. The aim of the invention is therefore to propose such a method. It consists in applying the correction function to D packets as described above but also in applying it to L packets numbered i, i+1, i+2, ..., i+ L-1. In this way, each data packet is used to compute two correction packets, and the correction rate in the presence of random errors increases significantly while maintaining good correction performance for series of data packets. The invention concerns a method for securing a data packet stream involving the following steps: arranging the packets to be sent in a matrix of D rows and L columns; to each row and each column of the matrix, applying an error correction function, the result of which is a correction packet; then sending the resulting correction packets in addition to the data packets. According to another aspect of the invention, the correction function is also applied to the row formed by correction packets resulting from the application of the correction function to each column of the matrix generating an additional correction packet. According to another aspect of the invention, the correction function is also applied to the column formed by correction packets resulting from the application of the correction function to each row of the matrix generating an additional correction packet. According to another aspect of the invention, the headers of the correction packets contain a field for distinguishing whether they have been computed on a row or a column of the matrix. According to another aspect of the invention, the correction packets are transmitted in the same stream as the data packets. According to another aspect of the invention, the correction packets are transmitted in a different stream from the one conveying the data packets. The invention also concerns a method for reconstructing lost packets in a data packet stream involving the following steps: arranging the received packets in a matrix of D rows and L columns, taking their sequence number into account; to each row and each column of the matrix containing at least one lost packet, applying a function for reconstructing missing packets using the received packets and the correction packet corresponding to the row or to the column. According to another aspect of the invention, the reconstruction function can also be applied to a row or a column of correction packets using an additional correction packet generated for this purpose. The invention also concerns a transmitter device for transmitting a data packet stream over a network containing computation means generating a correction packet by applying a correction function to a group of data packets and characterized in that the device contains means for applying this function to the rows and to the columns of a matrix formed by data packets. According to another aspect of the invention, the means applying the function do so according to the method disclosed in the patent. The invention also concerns a receiver device for receiving a data packet stream over a network containing means for computing lost packets as a function of the data packets and correction packets received, characterized in that the device has means for applying these computation means to the rows and to the columns of a matrix of data packets of the stream and to the correction packets corresponding to these rows and these columns. According to another aspect of the invention, the correction packets are generated according to one of the methods disclosed in the patent. The invention will be better understood, and other features and advantages will become apparent, on reading the following description which makes reference to the accompanying drawings in which: Figure 1 is a block-diagram representation of the application of the correction function to data packets. Figure 2 represents the structure of a correction packet placed in a packet according to the RTP protocol. Figure 3 represents the header of a correction packet according to RFC 2733. Figure 4 represents the header of a correction packet according to the example implementation of the invention described. Figure 5 represents the hardware architecture of a transmitter or receiver device according to the example implementation of the invention. Figure 6 represents the steps of the method for the transmitter. Figure 7 represents the steps of the method for the receiver. Figure 5 represents the internal architecture of a transmitter or receiver device 1 which contains ROM (Read-Only Memory) 4 enabling it to store programs and data, and RAM (Random Access Memory) 3 enabling it to load these programs in preparation for an execution by the processor 2. This device is connected to an IP network via a network interface 5 which means that it can transmit or receive the streams. These components communicate via an internal bus 6. The example implementation of the invention is in the context of digital data transfer in real time over an IP network. The data is transmitted, in this example implementation, in the form of a stream via the RTP (Real Time Protocol) protocol. As already mentioned, RFC 2733 describes a standard way to correct data packets transmitted by the RTF protocol. This method was the subject of an extension described in the Pro-MPEG forum's "Code of Practice" document. The example implementation of the invention which will now be described is an improvement that is compatible with this method. In this example implementation, the data packets are separated into groups of L*D packets. They are then arranged in a matrix of D rows and L columns, as Figure 1 shows. L and D are chosen according to the effectiveness desired. The way in which these parameters affect the effectiveness of the correction method will be described later. A correction function is then applied to each column of the matrix; the same correction function is also applied to each row of the matrix. This correction function can be an XOR, which would be the simplest to implement, but other functions are also possible such as the Reed-Solomon function or a Hamming code. These functions are more powerful but more costly in terms of computation. Whichever correction function is chosen, the result of applying this function to a group of packets, in this case a row or a column of the matrix, is a packet referred to as the correction packet. The method therefore results in L correction packets corresponding to the application of the function to each column and D correction packets corresponding to the application of the function to each row. It is also possible to add a correction packet computed by applying the correction function to the correction packets themselves, thus adding a second level of protection for correcting a loss of a correction packet. These packets must be transmitted to the data stream. This transmission can be performed in several ways. The correction packets can be transmitted in the same stream as the data packets, but they are then subject to the same transmission risks as the stream they are protecting. They can also be transmitted in a separate stream, thus increasing the resistance to errors. The method described by the Pro-MPEG forum consists in computing and sending the correction packets computed from the columns of the matrix. The method disclosed in the patent, insofar as it sends these same packets, remains compatible with the Pro-MPEG method. This is because receiver devices according to this standard will ignore the correction packets computed on the rows of the matrix and will therefore be able to operate in the same way with a transmitter according to the invention and with a transmitter according to the Pro-MPEG method. The structure of a packet, as described in the RFC, is represented in Figure 2. It consists of an RTP packet header, a description of which can be found in RFC 1889. This header is followed by an FEC header and then the actual correction packet. The structure of an FEC header according to RFC 2733 is represented in Figure 3. This header contains an SN base field which contains the lowest sequence number of the group of data packets used to construct the correction packet. The Length recovery field is used to determine the length of any data packet. It is constructed by applying the correction function to the group formed by the lengths of each data packet. The E field is used to indicate an extension to the header. The PT recovery field is obtained by applying the correction function to the data packet Type field. The Mask field is a 24-bit field for determining which data packets are used to construct the correction packet. If bit i is at 1, this means that packets N + i are used to construct the correction packet. N is the base stored in the SN base field. The TS recovery field is computed by applying the correction function to the timestamps of the data packets. It is used to reconstruct these timestamps. It is observed that through this mechanism, correction packets can onjy be computed on groups of a maximum of 24 data packets. To exceed this limit, an extension to this header is defined in the Pro-MPEG forum's "Code of Practice" document. This extension is represented in Figure 4, which shows the same header as previously, with the extension field set to 1. A new 2-bit extension field, E, is added which determines the use of the "Extension field" field. The Type field indicates which correction function is used: 0 for XOR, 1 for a Hamming code, 2 for a Reed-Solomon function. The Index field is used to order the FEC packets in the event that the result of the correction function exceeds the maximum size of a packet. This can be the case for complex correction functions. The Offset field determines the period chosen to select the data packets; it corresponds to the parameter L of the matrix. The NA (Number of Associated) field determines the number of associated data packets for the generation of the correction packet; this number corresponds to the parameter D of the matrix. The "Extension field" field is reserved for future use. The Mask field has become useless and is initialized by filling bytes having a value of 0. The modification brought about by the invention consists in separating the 2-bit E field into a new 1-bit E' field and a 1-bit D field which determines whether the correction packet in question is computed on the columns, in which case this D field is set to 0. If the correction packet is computed on a row, this D field is set to 1. It is observed that through this mechanism, a correction packet computed on a column remains completely identical to the same packet computed according to the Pro-MPEG forum method, while the additional packets computed on the rows can be identified by the presence of the D field set to 1, which amounts to setting to 1 the 2-bit extension field, E, if the packet is interpreted according to the Pro-MPEG method indicating that this packet is using an extension in the format defined by the forum.When packets are being received, the receiver will store the received packets in a buffer. The size of this buffer must be such that it can store at least L*D data packets and the L+D corresponding correction packets. The distinction between the data packets and the correction packets is made by the Type field of the RTP header as described in RFC 1889 that describes this protocol. It will be possible to identify the data packets among them through their sequence number and the correction packets among them through the SN base field of their header. Knowledge of the missing packets will then enable an attempt to be made to reconstruct them by using the correction function and the correction packets. For example, if the correction function used is the XOR function, it is possible to correct the loss of one data packet in a group of data packets used to generate a correction packet, using this correction packet. There will therefore be identified in the matrix of received packets, the rows and columns containing only one lost packet in order to reconstruct them with the corresponding correction packet. The correction can, for example, be carried out using the following method: Computation of a vector NLi of the number of missing data packets in row i. Computation of a vector NCi of the number of missing data packets in column i. - While there remains a missing packet in the matrix: o Find the minimum index p such that NLp = 1; • If such a p exists: • Find the column index q corresponding to the missing packet • Reconstruct the missing packet of index p and q by applying the XOR function to the other packets of the row and to the correction packet corresponding to this row • NLp = 0;NCq = NCq-1; • In the absence of such an index p find a minimum index q such that NCq = 1 • If such a q exists: o Find the column index p corresponding to the missing packet o Reconstruct the missing packet of index p and q by applying the XOR function to the other packets of the row and to the correction packet corresponding to this column o NLp = NLp-1;NCq = 0; • Else, the reconstruction method has failed. In the event of success, all the lost packets are reconstructed. Parameters L and D of the method must be chosen with consideration to the following information. First, the overhead generated by the correction packets which must be sent in addition to the data packets is 1/L+1/D; therefore choosing large values for L and D will minimize occupation of the bandwidth required for the use of the method. Furthermore, the method can be used to correct series of lost packets of at most L+1 adjoining packets, so the larger L will be, the more effective the method will be in correcting this type of error. The method is capable of correcting all random packet losses up to three lost packets and certain random packet losses up to D+L-1 lost packets. The symmetry of the problem leads to a maximum effectiveness on the random packet losses when L and D have the same value. Now, the method requires buffers in both the receiver and transmitter devices, having a minimum size of L*D packets. It is therefore observed that choosing large values for L and D is beneficial to maximize the effectiveness of the method but the choice is restricted by the size of the transmit and receive buffers of the devices. Another limitation can arise from the correction function computation time which can become long for a large number of packets and if a complex function is chosen. Statistical calculations can be used to compare the effectiveness of the method with that described by the Pro-MPEG forum. As expected, its effectiveness is observed to be markedly increased. For example, by using the XOR correction function and a matrix in which L and D have a value of 6, the method corrects 94.1 per cent of random losses of 6 data packets while the RFC method corrects only 2.4 per cent. The method therefore turns out to be much more effective in correcting random losses of data packets in the stream. The method is capable of correcting all losses of adjacent packets up to D+1 successive packets where the Pro-MPEG method corrects only up to D successive packets. It is therefore observed that the performance on the loss of adjacent data packets is slightly improved while the performance on random errors is greatly increased. It is clear, for the person skilled in the art, that the example implementation described here is not limiting. In particular, the invention can be used in the context of all protocols for transferring data by packets and not only in the context of RTP. It is also possible to use any correction function; the invention is not limited to the correction functions mentioned in the patent. WE CLAIM: 1. Method for securing a data packet stream transmitted by a transmitter device (1) comprising computation means (2) and means for applying said computation means (2) to each row and each column of a matrix, comprising a step of: arranging the packets to be sent in a packet matrix of D rows and L columns; characterized in that it further comprises following steps: generating a correction packet, for each row and each column of the matrix, by said computation means (2) by applying an error correction function to each row and each column of the matrix, said correction packet enabling the reconstruction of at least one lost packet in the row, respectively the column, to which the error correction function is applied; sending the resulting correction packets in addition to the data packets. 2. Method as claimed in claim 1, wherein the correction function is also applied to the row formed by correction packets resulting from the application of the correction function to each column of the matrix generating an additional correction packet 3. Method as claimed in claim 1, wherein the correction function is also applied to the column formed by correction packets resulting from the application of the correction function to each row of the matrix generating an additional correction packet. 4. Method as claimed in any one of the preceding claims, wherein the headers of the correction packets contain a field for distinguishing whether they have been computed on a row or a column of the matrix. 5. Method as claimed in any one of the preceding claims, wherein the correction packets are transmitted in the same stream as the data packets. 6. Method as claimed in any one of claims 1 to 4, wherein the correction packets are transmitted in a different stream from the one conveying the data packets. 7. Method for reconstructing lost packets in a data packet stream received by a receiver device (1) comprising means for applying computation means (2) to each row and each column of a matrix, which method is characterized in that it involves the following steps: arranging the received packets in a matrix of D rows and L columns, taking their sequence number into account; to each row and each column of the matrix containing at least one lost packet, applying a function for reconstructing missing packets using the received packets and the correction packet corresponding to the row or to the column by using said means for applying the computation means (2). 8. Method as claimed in claim 7, wherein the reconstruction function can also be applied to a row or a column of correction packets using an additional correction packet generated for this purpose. 9. Transmitter device (1) for transmitting a data packet stream over a network containing computation means (2) generating a correction packet by applying a correction function to a group of data packets and characterized in that the device contains means for applying this function to the rows and to the columns of a matrix formed by data packets, said correction packet enabling the reconstruction of at least one lost packet in the row, respectively the column, to which the error correction function is applied. 10. Device as claimed in claim 9, wherein the means applying the function do so according to a method described in one of claims 1 to 6. 11. Receiver device (1) for receiving a data packet stream over a network containing means (2) for computing lost packets as a function of the data packets and correction packets received, characterized in that the device has means for applying these computation means to the rows and to the columns of a matrix of data packets of the stream and to the correction packets corresponding to these rows and these columns. 12. Device as claimed in claim 11, wherein the correction packets are generated according to one of the methods described in claims 7 and 8.

Documents:

1646-DELNP-2006-Abstract (28-10-2009).pdf

1646-delnp-2006-abstract.pdf

1646-DELNP-2006-Claims (28-10-2009).pdf

1646-delnp-2006-claims.pdf

1646-DELNP-2006-Correspondence-Others (28-10-2009).pdf

1646-DELNP-2006-Correspondence-Others-(09-04-2010).pdf

1646-DELNP-2006-Correspondence-Others-(17-08-2006).pdf

1646-DELNP-2006-Correspondence-Others-(30-03-2010).pdf

1646-delnp-2006-correspondence-others.pdf

1646-DELNP-2006-Description (Complete) (28-10-2009).pdf

1646-delnp-2006-description(complete).pdf

1646-DELNP-2006-Drawings (28-10-2009).pdf

1646-delnp-2006-drawings.pdf

1646-DELNP-2006-Form-1 (28-10-2009).pdf

1646-delnp-2006-form-1.pdf

1646-delnp-2006-form-18-(17-08-2006).pdf

1646-DELNP-2006-Form-2 (28-10-2009).pdf

1646-delnp-2006-form-2.pdf

1646-DELNP-2006-Form-3 (28-10-2009).pdf

1646-delnp-2006-form-3.pdf

1646-delnp-2006-form-5.pdf

1646-DELNP-2006-GPA (28-10-2009).pdf

1646-delnp-2006-gpa.pdf

1646-delnp-2006-pct-105.pdf

1646-delnp-2006-pct-237.pdf

1646-delnp-2006-pct-304.pdf

1646-delnp-2006-pct-306.pdf

1646-delnp-2006-pct-409.pdf

1646-delnp-2006-pct-416.pdf

1646-DELNP-2006-Petition-137 (28-10-2009).pdf