WO2003096538A2

WO2003096538A2 - Method and device for recovering a code word from a received faulty code word, method and device for generating a code word, and transmission system

Info

Publication number: WO2003096538A2
Application number: PCT/EP2003/004758
Authority: WO
Inventors: Markus Kaindl; Joachim Hagenauer
Original assignee: Technische Universität München
Priority date: 2002-05-07
Filing date: 2003-05-06
Publication date: 2003-11-20
Also published as: AU2003226827A8; DE10220370A1; AU2003226827A1; WO2003096538A3

Abstract

The invention relates to a method and device for recovering a code word from a received faulty code word, according to which the faulty code word is firstly received (5100). The code word comprises code units and is derived from one or more information units by means of a redundancy code. Once the faulty code word is received, code units that are faulty or not present in the code word (5110) are determined and a de-dotting model is adapted (5112) in response to these code units. The received, faulty code word is then de-dotted (5114) based on the adapted de-dotting model thereby de-dotting the code unit that is faulty or not present in the code word.

Description

METHOD AND DEVICE FOR RECOVERING A CODE WORD FROM A RECEIVED, ERROR CODE WORD, METHOD AND DEVICE FOR GENERATING A CODE WORD, AND TRANSMISSION SYSTEM

description

The present invention relates to a method and apparatus for recovering code words from a received defective code word, and more particularly to a method and apparatus for recovering a code word transmitted over a lossy packet-oriented connection network and comprising a plurality of packets , Furthermore, the present invention relates to a method and to an apparatus for generating a plurality of packets for transmission over a packet-oriented connection network and to a transmission system. The present invention is particularly suitable for the transmission of multimedia data via a lossy channel, channels in particular being considered in which both packet losses and individual, randomly distributed bit errors occur in the data stream.

In the transmission of multimedia data, that is to say voice data, image data, video data and audio data, the trend in the development of future communication systems is moving away from connection-oriented transmission towards packet-oriented transmission. This means that a permanent connection is no longer switched exclusively for a connection between two communication partners, but that all participants use all lines together. To do this, it is necessary that the data stream of a subscriber is divided into packets and the divided packets are sent sequentially to the recipient. The individual packet independently searches for its way to the recipient in the network. The following effects can occur in such networks:

Packet losses, e.g. due to overloaded network nodes,

different transit times of the individual packets, since it is not guaranteed that all packets choose the same route in the network, and

- a received packet contains with high certainty no bit errors.

The packet-based transmission of multimedia data via a communication system, which consists of a line-bound core network (fixed network) and a radio network (mobile network), is considered below. The core network bridges long distances, while the radio network ensures that a mobile subscriber can be reached wirelessly on the "last mile".

A concrete example of such a transmission system would be e.g. B. the combination of the Internet with a modern mobile radio system, e.g. B. GSM-GPRS / UMTS, which is often referred to as mobile Internet.

The following effects occur when transmitting multimedia data over such a data network:

Packet losses in the core network, since nodes of the core network may be overloaded or individual packets may not arrive at the recipient in time, and

Bit error in the transmission through the radio network.

In such a communication system, the receiver is thus confronted with a data stream in which packets may be missing on the one hand and packets received on the other hand may contain individual bit errors. When transmitting multimedia data via a system of the type mentioned above, there are various methods in the prior art which can essentially be divided into the five classes explained in more detail below.

I. Error concealment at the recipient:

Methods of this class aim to recognize the transmission errors on the receiver side and to conceal them by means of a suitable signal interpolation. Although these methods deliver very good results with low bit error rates or individual packet losses, they usually fail with higher bit error rates and with successive packet losses.

An example of such a method is described, for example, in ITU-T Recommendation G. 111 - Appendix 1, "A high quality low-complex algorithm for packet loss concealment with G. 711", ITU-U, 1988

II. Interleaving / Multiplexing at the Sender: Methods of this class spread the data of a packet over several packets (interleaving) or bundle the data streams of several participants (multiplexing) before transmission. This is associated with a delayed transmission of the data

(Buffering). This can be especially the case with dialog-oriented

Data transmission, e.g. B. Telephony, too intolerable

Lead to delays. These methods are particularly suitable for transmission via packet loss channels, since a packet lost during transmission (bundle errors) leads to individual, distributed bit errors in the data stream, which are then well masked using the error concealment method described above or corrected using the channel coding method described below can.

An example of such a method is given by SV Anderson, et al. a. , "Mul tiplexed Predictive Coding of Speech" in Proc. Of ICASSP 2001, Sal t Lake City, Utah, May 2001, Vol. III, Be described 741 to 744 th

III. Channel coding: Methods of this class specifically insert redundancy (channel coding) into the data stream at the transmitter, which can be used to detect and correct transmission errors at the receiver.

An example of this method is e.g. by M. Luby, u. a. "Practical Loss Resilient Codes" in Proc. 9th Annual ACM-SIAM Symposium on Discrete Algorithms, January 1998.

IV. Multiple Descriptions:

Methods of this class generate several, usually two, descriptions of the same from a data packet. If only a part of the descriptions is now received, decoding with reduced quality can take place. If all descriptions are received, the decoding takes place in the original quality. These methods are particularly suitable for transmission via packet loss channels, but have so far found little use.

Such a method is described, for example, by W. Jiang, et al. a. "Multiple description speech coding for robust communication over lossy packet networks", described in International Conference on Multimedia and Expo, New York, NY, USA, August 2000, Vol. II, pages 444-447.

V. ARQ procedure:

Methods of this class recognize lost packets on the receiver side and request them again from the sender. This requires a return channel from the receiver to the transmitter. Another disadvantage of these methods is that there is a considerable delay until all packets have actually been received. An example of such a method is by J. Hagenauer, "Rate-Compatible Punctured Convolutional Codes (RCPC Codes) and their Applications", IEEE Transactions on Communications, Vol. 36, No. 4, pp. 389-400, April 1988.

Depending on the specific application, it is also possible to combine the processes described in more detail above.

The disadvantage of the above-described methods is that they can either no longer provide the necessary correction at high bit error rates, lead to intolerable delays or do not allow secure transmission.

On the basis of this prior art, the present invention is based on the object of making it possible to transmit data with improved quality in a simplified manner in a simplified manner.

This object is achieved by a method according to claim 1, a device according to claim 11, a method according to claim 21, a device according to claim 22 and a transmission system according to claim 25.

The present invention provides a method for recovering a code word from a received, faulty code word, wherein a code word comprises code units and is derived from one or more information units by means of a redundancy code, with the following steps:

(a) receiving the code word;

(b) determining a defective or non-existent code unit in the code word; (c) in response to the defective or non-existing code unit determined in step (b), setting a depuncturing pattern; and

(d) Depuncturing the received, defective code word by means of the set depuncturing pattern, so that the defective or non-existing code unit is depunctured in the code word.

The present invention also provides a device for recovering a code word from a received, incorrect code word, wherein a code word comprises a plurality of code units and was generated by means of a redundancy code

a signal processing unit which receives the faulty code word, which determines a faulty or non-existent code unit in the code word, which sets a de-puncturing pattern in response to the faulty or non-existent code unit, and which punctures the received code word based on the set de-puncturing pattern, so that the defective or non-existent code unit is depunctured in the code word.

The present invention further provides a method for generating a code word, which contains code units, which are generated based on information units by a redundancy code, for transmission over a lossy connection network, with the following steps:

(a) receiving the information units;

(b) coding each information unit to generate a plurality of code units for each information unit; and

(c) arranging the generated code units in the code word in such a way that a loss of one code unit or multiple rer code units in the code word leads to a code unit sequence which corresponds to a puncturing pattern applicable to the original code word, which is also known to a recipient of the code word.

The present invention further provides a device for generating a code word, which contains code units which are generated based on information units by a redundancy code, for transmission via a lossy connection network, comprising:

an input for receiving the information units;

an encoder for encoding each information unit to generate a plurality of code units for each information unit; and

a processing device for arranging the generated code units in the code word such that a loss of one or more code units in the code word leads to a code unit sequence which corresponds to a puncturing pattern applicable to the original code word, which is also known to a recipient of the code word ,

According to a preferred embodiment, a plurality of packets are generated which contain code units which are generated based on information units by a redundancy code and which are transmitted via a packet-oriented connection network. Here, the information units are received and each of the received information units is encoded using a plurality of generators, one information unit being applied to each of the generators and each of the generators generating a code unit for an adjacent information unit. The code units are assigned to a plurality of packets, with code units generated by the same generators being assigned to the same packets. ordered and the generated packages are provided.

The present invention also provides a transmission system for transmitting units of information

a transmitter which comprises a device for generating a code word or a device which operates according to the inventive method for generating a code word;

a transmission medium that receives and transmits the code words generated by the transmitter; and

a receiver which receives the code words transmitted by the transmission medium, the receiver comprising a device for recovering code words according to the present invention, or a device which operates according to the method according to the invention for recovering the code words.

The present invention is basically to be assigned to the channel coding method described above and is based on punctured channel codes, which are also described by J. Hagenauer, "Rat-Compatible Punctured Convolutional Codes (RCPC Codes) and their Applications", IEEE Transactions on Communications, Vol. 36 4, pp. 389-400, April 1988. For further optimization, it is also possible to combine the method according to the invention with one or more of the transmission methods described above, in particular the error concealment method, the interleaving / multiplexing method. Procedure and the ARQ procedure come into consideration.

With regard to puncturing, it should be pointed out that this means the systematic removal of bits before the transmission of a data stream. Does the decoder know which bits on the transmitter side te have been removed (punctured), the decoder can still decode the data stream. The performance of a punctured channel code of rate r, which was obtained from a mother code at rate R _M , is only slightly worse than a real channel code of rate r. By puncturing, a code of any rate r can be generated very easily from a mother code of rate R _M , where 1>r> R _M applies.

In conventional systems, the puncturing of the mother code always takes place immediately after the channel encoder on the transmitter side according to a defined puncturing pattern which is known to both the transmitter and the receiver. The present invention leaves this approach and according to the invention there is no puncturing of the channel code on the transmitter side. This means that basically all bits of the generated mother code are sent. In contrast to the prior art, according to the invention the puncturing technique is not used on the transmitter side, but is shifted to the receiver side, the puncturing actually occurring, i.e. the discarding of certain bits, being caused by transmission losses, i.e. by the fact that during transmission a Bitstream single or multiple bits are lost, i.e. they do not arrive at the receiver. This can occur, for example, when transmitting via a packet loss channel.

The essence of the present invention is therefore that the puncturing is virtually shifted from the transmitter to the receiver, the actual puncturing being effected by losses in the transmission network. The present invention preferably relates to the use of convolutional codes, since the puncturing technique is particularly suitable for this. However, the present invention is not limited to this.

This packet loss channel is considered as an example for the further explanation, although it is obvious that that the present invention is not limited to such a packet loss channel, but can be applied to any lossy transmission channel and the associated transmission methods.

If one now looks at the packet loss channel and a packet is actually lost during transmission via the same, then on the receiver side, which comprises, for example, an FEC decoder (FEC = Forward Error Correcting), it is assumed that all bits of the lost Packages were virtually punctured at the transmitter. However, since the packet losses are purely coincidental and can hardly be predicted by the recipient, the recipient must be informed whether a particular package has been lost or not. However, this is only a small piece of additional information that can be signaled very easily via a control channel or an inband. The same can be achieved in that those packets which have not arrived at the FEC decoder after a certain time are declared as punctured (ti e out).

The method according to the invention can thus very efficiently correct both packet losses and individual, randomly distributed bit errors in one step.

In principle, the method according to the invention can be applied to any channel coding method, e.g. B. convolutional codes, turbo codes, etc. A decoder with a soft input or at least one erasure decoder (“erasure”) should preferably be available for the corresponding channel code.

The method according to the invention thus leads to a substantially secure transmission of the data in comparison with conventional, commercial methods as described above, which, for example, in the transmission of multimedia data, depending on the application, is significantly improved. voice quality, audio quality, picture quality or video quality.

Previous FEC procedures were always limited to a special channel. So z. B. for transmission via packet loss networks especially Reed-Solomon codes, which, however, cannot be decoded soft-in / soft-out, or only at considerable expense. In contrast, due to their code properties, convolutional codes are rather unsuitable for correcting packet losses, but they can be decoded very well soft-in / soft-out. Convolutional codes are particularly suitable for transmission over mobile radio networks in which individual, distributed bit errors occur, while Reed-Solomon codes or block codes are rather unsuitable.

According to a preferred exemplary embodiment, the method according to the invention processes the data before sending it in such a way that a packet loss leads to optimal puncturing of the convolutional code. This is achieved by optimally distributing the encoded data over a plurality of packets. This ensures that a packet loss that would normally not have been correctable by a convolutional code remains correctable up to a certain packet error rate. The preprocessing at the transmitter can guarantee at the receiver that M 'packets sent can be lost, M' <M, and an error-free decoding result can nevertheless be achieved. M 'depends on the code rate N and the code used.

According to a preferred exemplary embodiment, the method according to the invention is used on a combination of fixed network and mobile radio network, but the method according to the invention also works according to another exemplary embodiment for pure packet loss networks or pure mobile radio networks. However, the use in the combination of the two networks mentioned above is particularly suitable, since here both Packet losses of the fixed network and the bit errors of the mobile radio network can be corrected in one step with the same code, which, as already explained above, was not possible in the prior art.

The present invention has the following advantages:

both packet loss and individual errors can be handled,

the invention is based on channel codes punctured by packet loss,

in the event of a packet loss, puncturing occurs virtually on the receiving end,

no back channel is required

there is no additional delay in the data stream,

any punctuable methods can be used as channel code,

a coding can optionally be used on a base station in a configuration of the transmission network as a combination of a fixed network and a mobile network, without the recipient having to be aware of this,

the sender can vary the number and length of packets as desired without the recipient needing to be aware of it, so that the system can be adapted very flexibly to the transmission quality and capacity of the individual subnets

the recipient only needs to be informed that a packet has been lost, due to the structure, all adjustments to the system, such as B. mode, rate, carrier, are not signaled to the receiver, so that this adaptation can be done flexibly and depending on the load,

Adjustments to the cellular network do not have to be signaled to the transmitter, such as. B. discarding packets because the cellular connection has fewer errors,

no real, computationally complex transcoding has to take place at the base station.

Further developments of the present invention are defined in the subclaims.

Preferred exemplary embodiments of the present invention are explained in more detail below with reference to the accompanying drawings. Show it:

1 shows a flowchart of the method according to the invention in accordance with a preferred exemplary embodiment;

2 shows a transmission system according to a preferred exemplary embodiment of the invention with the receiver according to the invention;

FIG. 3 shows the behavior of the transmission system shown in FIG. 2 without re-coding the code words received in the base station;

FIG. 4 shows the behavior of the transmission system shown in FIG. 2 with re-coding of the code words received in the base station; 5 shows a conventional transmission system for packet-oriented transmission between transmitter and receiver;

Fig. 6 illustrates the behavior of the conventional transmission system shown in Fig. 5; and

Fig. 7 is a tabular comparison of the behavior of the conventional transmission system and the transmission system according to the invention.

In the following description of the preferred exemplary embodiments, similar or equivalent elements are provided with the same reference symbols in the individual figures.

A preferred exemplary embodiment of the method according to the invention for recovering a code word is explained in more detail with reference to FIG. 1. This method assumes that the data stream transmitted from a transmitter to a receiver comprises a plurality of code words, wherein individual code words have a plurality of packets, with a plurality of code units, e.g. B. bits are included, which were generated on the transmitter side by a suitable coding method and were distributed to individual packets according to a predetermined scheme.

In step S100 the code words are received and in step S102 it is checked whether all packets of the code word have been received. If this is the case, the method goes to step S104, in which the received code word is decoded so as to recover the information units on which the coding in the transmitter was based, that is to say the transmitted multimedia data. Then the decoded data for further processing, eg. B. display, manipulation, etc. forwarded. If it is determined in step S102 that not all packets of the code word have been received, the method proceeds to step S110, in which it is determined for the received code word how many packets have not been received and at which position / position they have been arranged in the code word are. Based on this information, it is determined in the following step S112 that the received code word, at least from the point of view of the receiver, has obviously been punctured at the transmitter in accordance with a predetermined puncturing pattern. Based on this information, the method according to the invention selects a suitable depuncturing pattern in step S112, which is used in subsequent step S114 to depuncture the code word, as a result of which the code word is completed again. After the code word has been completed, the method goes to step S104, in which the required decoding of the code word can be carried out.

According to a preferred embodiment, the decision S102 made regarding the number of packets received is based on another information signal which indicates whether or not a particular packet from a set of transmitted packets has been received. Alternatively, a period of time during which the packets of a code word are expected can be monitored, and in the event that one or more packets are received outside this period of time, these are rated as not received.

It is possible to decode a received code word if a packet was received without errors. However, it is not generally necessary for a packet to be received correctly. There may be situations in which several packets have been received, each of which contains part of the data it contains. In this case, error-free decoding would also be possible. A preferred exemplary embodiment of a transmission system which uses the method according to the invention for retrieving a code word and thus enables the advantages described above for the transmission of data is explained in more detail below with reference to FIG. 2. In the exemplary embodiment shown in FIG. 2, the transmission of multimedia data to a mobile receiver via a combined packet loss network and mobile radio network is described.

The multimedia data to be transmitted are from various multimedia sources, e.g. B. a camera 100a or a microphone 100b, provided to a multimedia encoder 102. The multimedia encoder 102 generates a plurality of information units based on the signals present from the multimedia sources 100a, 100b, which are forwarded to a channel coding unit 106 via a connection 104. The channel coding unit 106 comprises an FEC encoder 108, a packet distribution unit 110 and a multiplexer 112. The FEC encoder 108 receives the information units at its input via the connection 104 and generates N code units, e.g. B. Bits that are forwarded to the packet distribution unit 110 via the connections shown in FIG. 2. The packet distribution unit 110 distributes the received N code units to M packets, preferably M> N. The M packets are passed on to the multiplexer 112, which then provides them serially at its output. The output of the channel coding unit 106, ie the output of the multiplexer 112, provides the packets generated to a packet loss network, that is to say a packet-oriented connection network 114, which serially transmits the packets to a base station 116. The base station 116 comprises a demultiplexer 118 which provides the M packets transmitted via the network 114 in parallel at an output thereof. Optionally, a new coding unit 120 can be provided in the base station 116, in which case that at the output of the demultiplexer 118 parallel M packets can be provided in parallel to the re-encoding unit 120.

The base station transmits the received M packets, either the M packets received directly at the base station or the M packets newly coded in the base station, via a radio network shown only schematically at 122 in FIG. 2, to an FEC decoder 124 in the mobile receiver. The FEC decoder 124 generates the information units initially provided to the channel coding unit 106 from the received M packets and transmits them via the connection 126 to a multimedia decoder 128 which decodes the received information units and corresponding multimedia devices which are suitable for reproducing the corresponding multimedia data. provides e.g. B. the screen 130a or the speaker 130b.

The transmission system shown by way of example in FIG. 2 further comprises a controller 132 for the packet loss network and a controller 134 for the mobile network. As can be seen from FIG. 2, the controller 134 for the mobile network receives the DQI signal (DQI = Downlink Quality Indication) from the FEC decoder 124 and issues a carrier command to the base station 116. As indicated by the connection 136, the controller 132 for the fixed network and the controller 134 for the mobile network exchange information which will be discussed later. The controller 132 for the fixed network receives the parameter PER from the packet loss network, that is to say the packet error rate (PER = Packet Error Rate), that is to say the rate at which packets are lost on the network 114. Based on the received information, the controller for the fixed network 132 issues the channel coding unit 106 with the rate command, which is used to determine the rate at which the packets are sent to the packet loss network 114. The controller for the landline 132 also issues a mode command to the multimedia encoder 102 to determine a mode thereof. In a further preferred exemplary embodiment, as is also shown in FIG. 2, provision can be made to provide an optional return channel between the FEC decoder 124 and the channel coding unit 106. Provision can also be made to forward the information provided to the base station 116 about the receipt of a faulty packet to the FEC decoder 124, as indicated by the optional connection “BFI 1-M” (BFI = Bad Frame Indicator).

The functionality of the transmission system according to the invention according to the illustrated embodiment is explained in more detail below with reference to FIG. 2.

As mentioned, FIG. 2 shows the transmission system according to a preferred exemplary embodiment in a possible application scenario of the present invention. In the embodiment shown in FIG. 2, the data sources, e.g. B. digitized source data provided by the camera 100a or the microphone 100b, processed by the multimedia encoder 102. The original data is compressed. In general, it is assumed that the degree of compression on the multimedia encoder 102 can be adjusted, so that the data rate at the output of the multimedia encoder can be adjusted in several stages by means of the mode command which is present at a control input of the multimedia encoder 102 , The compressed data is then provided to the channel coding unit 106 via the connection 104. Here the data is processed by means of the channel encoder or FEC encoder 108, which has a rate of 1 / N. This means that 108 bits by L coded bits are generated from L bits at the input of the FEC encoder at the output thereof. The rate of the FEC encoder 108 is set by means of the rate command which is present at a control input of the FEC encoder 108. The coded bits are now the Packet distribution unit 110 is provided which divides the encoded bits into M packets.

Since a packet loss on the decoder side is interpreted as a puncturing of the convolutional code, it is necessary to distribute the N times L coded bits over the M packets in such a way that a packet loss during transmission over the packet-oriented network 114 to one at the decoder received channel code, which is optimally punctured for this. In other words, it is necessary to generate at the output of the channel coding unit 106 a data stream to be transmitted to the decoder in the form of a code word, a loss of a packet in the code word and in the decoder being recognized as a punctured code word, with regard to the Arrangement of the packets in the code word can be used using known methods for puncturing without actual puncturing naturally taking place.

The so-called free distance d _fr ee, which the punctured code has, and its distance spectrum are preferably used to assess puncturing patterns with regard to their “optimality”. The distance spectrum consists of two components ad and Cd. Here, a _d gives the number the paths in the trellis that build a Hamming distance d to the null word and C _d the sum of the bit errors on these paths of distance d.

Both the largest possible free distance d _fr e _e and the smallest possible values a _d and c _d are considered optimal. See also, for example, J. Hagenauer: "Rate Compatible Punctured Convolutional Codes and their Applications".

The problem with this is that the optimal puncturing pattern can usually only be determined using complex computer simulations. In the application used here, the puncturing patterns depend on the code, the code rate, the packet length and the number of lost packets. at Convolutional codes are preferred to distribute the punctured bits as evenly as possible over a block.

Examples of the optimal puncturing patterns mentioned above are from J. Hagenauer, "Rate-Compatible Punctured Convolutional Codes (RCPC Codes) and their Applications", IEEE Transactions on Communications, Vol. 36, No. 4, p . 389-400, April 1988; by J. Hagenauer, et al., "Channel Coding and Transmission Aspects for Wireless Applications", IEEE Special Issue Video Transmission gor Mobile Multimedia Applications, Vol. 87, pages 1764-1777. October 1999, and by C. Weiß, "Error Correcting with Tail-Biting Convolutional Codes", dissertation, TU Munich, December 2001.

After the bits are distributed among the M packets, the M packets are sequentially sent over the packet loss channel 114.

By returning the packet error rate (PER) of the packet loss network 114 to the network control unit 132, it is possible to adaptively set both the compression of the multimedia encoder 102 and the rate of the FEC encoder 108. This makes it possible to adaptively adapt both the number of packets and the length of the packet to the current packet loss probability of the network.

The received data packets are transmitted at the base station (BS) 116 via the carriers of a mobile radio network 122. The only additional information that has to be transmitted to the receiver is the signaling BFI _X (Bad Frame Indication = display of a bad frame), which indicates whether a certain packet x has arrived at the base station (in time) and has thus been transmitted via the radio channel is or not. A control or control channel or in-band signaling can be used for this. The same can be achieved if the packets that have not arrived after a certain time (“time out ") in the FEC decoder 124 are declared as dotted.

In addition, in another exemplary embodiment, decoding with subsequent re-coding 120 (optional re-coding) can be carried out optionally on the base station 116 before transmission over the radio channel 112. As a result, the packet losses occurring in the packet loss network 114 are corrected before the radio transmission. This becomes a clear gain in particular if the bit error probability on radio channel 122 is high.

By returning the transmission quality DQI on the radio channel to the network control unit 134 of the mobile radio system, it is possible to adaptively adapt the number of carriers of the mobile radio system used to the current channel state of the radio channel. An exchange of feedback information between the two network control units 132 and 134 serves to optimize the transmission quality of the overall system. If an optional return channel (ARQ) is available, incremental redundancy can also be requested if there is a high probability of packet loss by retransmitting some of the M sub-packets.

The received data are then decoded in the FEC decoder 124 at the receiver end and forwarded to the multi-ediadecoder 128, which decompresses them.

Although a preferred exemplary embodiment of the application of the method according to the invention has been explained above, it should be pointed out that the method according to the invention can in principle be used in any communication system, although it is particularly suitable for systems in which the transmission characteristic is both packet loss and individual , randomly distributed bit errors in the corresponding data stream. Voice transmission over the Internet and GSM-GPRS, voice / audio / image / video transmission over the Internet and UMTS, voice / audio / image / video transmission over the Internet and GSM are particularly suitable for the present invention. EDGE viewed. Furthermore, when using concatenated channel codes, data can be transferred very efficiently via the system described above (file transfer).

The transmission properties and the behavior of the transmission system described with reference to FIG. 2 according to a preferred implementation of the same are explained in more detail below. For the implementation of the transmission system from FIG. 2, the example of the transmission of voice over the mobile Internet was considered. The term “mobile Internet” means the combined network described above, consisting of a packet-oriented network part and an additional radio channel connection section. As already mentioned above, the transmission system opens up a very flexible adaptation of the transmission characteristics depending on the losses on the packet-oriented network and As mentioned, the packet loss at the receiver is considered as puncturing the convolutional code, so the only additional information to be transmitted to the decoder is the bad frame indication mentioned above.

The multimedia encoder 102 encodes the speech based on the AMR speech codec, which is based on the CELP technology, which enables up to eight rates between 4.75 kbit / s to 12.2 kbit / s. The subjectively pure voice quality of the 12.2 kbit / s mode is comparable to the G.711 codec. In addition, a 6-bit CRC is applied to the most important bits. This CRC is used on the speech decoder for the detection of defective frames of a speech frame and controls the error coverage erroneous language frames. This leads to a total bit rate of 5.05 kbit / s to 12.5 kbit / s after the coding of the speech by the encoder 102.

In addition, the encoder 102 can also pass on source significance information (SSI) to the channel coding unit 106, which is then used, for example, for the unequal error protection (UEP) of the voice data. Subsequently, a recursive systematic convolutional code (RSC = RSC = Recursive Systematic Convolutional) with the rate 1 / N is applied to the data 104 output by the encoder 102 in the encoder 108. After a packet loss on the receiver side is regarded as puncturing the convolutional code, the packet formation of the coded bits should preferably be carried out according to an optimal puncturing pattern of the generated code.

In the exemplary embodiment shown, it is assumed that the packet formation is carried out in such a way that all bits which come from a specific generator polynomial G _X (D) are assigned to the same packet x. Therefore, at the output of the packet distribution unit 110, there is a packet that contains the systematic bits and Nl different packets, each of which contains the bits that were generated by a specific generator polynomial.

The N packets are then serially transmitted over the lossy packet-oriented network 114. By returning the PER signal, the speech coding mode and the rate of the RSC code can be adjusted adaptively, and consequently the number and length of the packets transmitted can be set.

The received packets are transmitted at the base station 116 via the carriers i to N _{N of} the wireless network or radio network 122. The only information that has to be transmitted additionally is with this version Example of the bad frame indicator (BFIx), which indicates whether a specific packet x = 1, ..., N arrived at the base station 116 in good time and was consequently transmitted via the wireless network or not. In addition, the optional re-encoding described above can be performed at base station 116 prior to transmitting the received packets. This is very helpful in dealing with the loss of packets on the landline 114, which can improve the overall behavior of the system.

Feedback of the DQI signal from decoder 124 to the base station via a control unit of the mobile radio network can be used to adaptively adapt the number of carriers used for the wireless connection.

The received packets are decoded at the receiver end, that is to say in the decoder 124, and the decoder 128 is designed as a speech decoder which receives the decoded packets. It should be noted that it is sufficient if only one random packet from the N packets, which is received without errors, is required to enable error-free decoding. If any remaining redundancy remains in the speech-encoded data, iterative source and channel decoding can alternatively or additionally be used.

Based on the implementation of the transmission system explained in FIG. 2 just described, the method according to the invention was analyzed. The number of packets N was chosen in such a way that the packet distribution is comparable both in the packet according to the invention, ie also in a reference system, which will be described later. The delay of the two systems is also the same. In order to have a suitable comparison between the two systems, all feedback information and adaptations in the system according to the invention have been neglected for the coding experiment. The AMR mode MR122 was used as the speech encoder, with a fixed rate of 12.2 kbit / s. In addition, the 6-bit CRC described above was applied to the most important bits, resulting in a total bit rate of 12.5 kbit / s after the voice coding. The source significance information mentioned above was not used in this example. A code with the rate 1/5 was selected as the recursive systematic convolutional code, with the following generator polynomials:

As a result, five packets were generated from one speech frame and sent over the packet-oriented network. The overall rate after packet distribution was 62.5 kbit / s.

A simple model of a lossy packet-oriented network was assumed for the transmission channel, this simple model assuming statistically independent packet losses, and for the wireless connection additive channels with white distributed noise (AWGN).

The spectral distance (SD = SD = Spectral Distance) was assumed as the measured value for the speech quality. a. "Distorsion Measures for Speech Processing" is described in IEEE Transactions on Acoustics, Speech and Signal Processing, Vol. 28, No. 4, pages 367 to 376, August 1980.

d (S _x (ω), S _y (ω)) = \ \ nS _x (ω) - \ nS _y (ω) \ ² dω

Here S _x (ω) and S _y (ω) are the amplitude spectra of the original, error-free transmitted system and the sig- output of the speech decoder. W denotes the signal bandwidth. Based on this assumption, only distortions due to the transmission are considered, distortions due to the speech coding are neglected. In the above equation, a small value d indicates little or little distortion of the signal transmitted without error, and consequently high speech quality. A high value of d indicates high distortion and consequently poor speech quality.

FIG. 3 shows the behavior of the system described in more detail in FIG. 2 according to the implementation mentioned above, wherein no re-coding was carried out in the base station 116 here. The spectral distance SD is plotted along the z axis as a measure of the voice quality, the packet error rate PER occurring on the network 114 is plotted along the x axis and the signal-to-noise ratio is plotted on the mobile network along the y axis ,

As can be seen from FIG. 3, packet losses on the packet-oriented network essentially do not influence the quality over a large range of packet error rates, slight distortions only occurring at high packet error rates of greater than 10%. Furthermore, the RSC code is able to correct most errors caused by the wireless connection. The speech quality only decreases at low signal-to-noise ratios (E _s / N ₀ <-1 dB). However, it should be noted that even in the worst or worst case, namely if PER = 0.25 and E _s / N ₀ is -4 dB, the transmitted speech or the transmitted speech signal is still understandable.

The behavior of the transmission system described above will now be described with reference to FIG. 4 if the optional re-coding is additionally carried out in the base station. As can be seen, this additional Neuco- The aim is to avoid all packet losses when transmitting over the packet-oriented network, only the distortions due to the transmission over the wireless network remain.

A reference system is explained in more detail below with reference to FIG. 5, which also transmits encoded voice data to a mobile subscriber via a lossy packet-oriented network and via a wireless connection, but uses a conventional approach to correcting packet losses, namely signal interpolation. 5, an encoder 202 is shown which, via a connection 204, forwards the encoded data to a packet distribution unit 210, which distributes the received data to a multiplicity of packets which are then transmitted to a base station 216 via the network 214. The received packets are transmitted to the mobile receiver via a radio channel 220, the received packets being demultiplexed in the demultiplexer 224 and passed on to the decoder 228. In the reference system shown in FIG. 5, the G.711 codec was chosen as the speech codec and the encoded data stream was distributed over N packets in such a way that adjacent samples were introduced into adjacent packets. Consequently, a packet loss leads to a distribution of the sample signal losses and not to a complete loss of sample values. A packet loss can then be concealed at the decoder simply by interpolating between the distributed sampling losses. The packets are transmitted via the same mobile Internet channel as in the exemplary embodiment of the present invention described above.

FIG. 6 shows the behavior of the transmission system shown in FIG. 5, and as can already be seen from a superficial comparison of FIGS. 3 and 4 with FIG. 6, the behavior of the reference system has a significantly poorer speech quality Areas in which the system according to the invention is still used for production language in the transmission. As can be seen from FIG. 6, the system suffers greatly from the non-coded transmission of the voice data via the wireless connection. But even with low bit error rates or high signal-to-noise ratios, the system is still very susceptible to transmission errors as far as the voice quality is concerned. If one now looks at the packet error rate of the lossy packet-oriented network, it can be seen that the reference system is also very sensitive to packet losses. Thus, a packet loss rate of 10 ^-3 already leads to a decrease in the speech quality, wherein, starting from a packet error rate of 10 ^-3 with increasing packet error rate, the transmission quality is getting worse. Compared with the system according to the present invention without re-coding, it can be ascertained that in the reference system, even with an increasing packet error rate from 10 ^"3, a significant decrease in the speech quality can be determined, which does not occur in the system according to the invention.

In order to verify the results obtained on the basis of FIGS. 3, 4 and 6, informal hearing tests were carried out, the parameter SD being used as a measure of the speech quality. The G.711 reference system (FIG. 6) was compared with the transmission system which implements the method according to the invention for recovering faulty code words (without re-coding), see FIG. 2. The table in FIG. 7 shows for the two systems what percentage of the test listeners preferred which system. The superiority of the new system with the approach according to the invention for recovering faulty code words is readily apparent, and based on a comparison of the signal-to-noise ratios and packet error values shown in FIG. 7 with FIGS. 3 and 6 is not surprising.

1 to 7, preferred exemplary embodiments of the present invention have been described, in particular In particular, the implementation of the method and the device according to the invention in a transmission system is described, which represents a combination of a radio network and a fixed network. The present invention is not limited to this application, but can be applied in general to any transmission networks, fixed networks or mobile radio networks. Furthermore, the present invention is not limited to the RSC coding described above, but any type of coding that inserts redundancy into the code words to be transmitted can be used with the method according to the invention, provided that the available code can be punctured (not on the Transmitter) so that losses in transmission can be recognized by the receiver as puncturing.

In addition, the present invention can also be applied to systems in which, in order to increase the transmission capacity, the generated codes are punctured, with an additional loss in the transmission being regarded as additional puncturing by the receiver.

The loss of the packets or the code units can also be caused in the receiver and / or after generation in the transmitter, in addition to the transmission.

Claims

claims

Method for recovering a code word from a received, faulty code word, a code word comprising code units and being derived from one or more information units by means of a redundancy code, with the following steps:

(a) receiving (S100) the code word;

(b) determining (S110) an incorrect or non-existent code unit in the code word;

(c) in response to the defective or non-existent code unit determined in step (b), setting (S112) a deposition pattern; and

(d) depuncturing (S114) the received, defective code word by means of the set depuncturing pattern, so that the defective or non-existing code unit is depunctured in the code word.

l. A method according to claim 1 with the following step:

(e) decoding (S104) the dotted code word to recover the one or more information items.

3. The method according to claim 1 or 2, wherein a defective code unit or a loss of a code unit in the transmission channel (114) is caused after the generation of the code word.

4. The method according to claim 3, wherein a defective code unit or a loss of a code unit in a transmitter (102, 110) after the generation of the code word and / or when the code word is transmitted to a receiver (116).

5. The method according to any one of claims 1 to 4, wherein a code unit is a packet with a plurality of bits, so that the received code word comprises a plurality of packets, wherein in step (b) it is determined whether a packet of the code word at transmission of the code word via a packet loss network (114) has been lost.

6. The method according to claim 4 or 5, wherein the depuncturing pattern is set depending on the number of lost packets and depending on their position in the code word.

7. The method according to any one of claims 4 to 6, wherein the step of (b) comprises receiving an information signal (BFI) indicating that a packet has been lost.

8. The method according to any one of claims 4 to 6, wherein step (b) comprises monitoring a period of time during which a packet has to arrive, a loss of the packet being determined when the period of time has elapsed without the packet having arrived ,

9. The method according to any one of claims 1 to 8, in which the received code word is transmitted to a mobile receiver via a radio network (112) and in which steps (b), (c) and (d) are carried out in the receiver become.

10. The method of claim 9, wherein prior to transmitting the received code word to the mobile receiver, the received code word is re-encoded (122).

11. Device for recovering a code word from a received, defective code word, wherein a code word comprises a plurality of code units and was generated by means of a redundancy code with

a signal processing device (116; 124) which receives the defective code word, which determines a defective or non-existent code unit in the code word, which sets a depuncturing pattern in response to the defective or non-existent code unit, and which receives the received one based on the set defuncturing pattern Depunits the codeword so that the defective or non-existing code unit is depunctured in the codeword.

12. The apparatus according to claim 11 with a decoder

(124, 128) who receives the depunctured code word and retrieves the one or more information units from the depunctured code word.

13. The apparatus of claim 11 or 12, wherein the transmission channel (114) causes the incorrect code word after the generation of the code word.

14. The apparatus of claim 13, wherein a transmitter (102, 110) and / or a transmission medium (114) between the transmitter (102, 110) and a receiver (116; 124, 128) causes the incorrect code word.

15. The apparatus according to any one of claims 11 to 14, wherein a code unit is a packet having a plurality of bits so that the received code word comprises a plurality of packets, the signal processing unit determining whether a packet of the code word is transmitted during the transmission of the code word lost via a packet loss network (114).

16. The apparatus of claim 14 or 15, wherein the signal processing unit sets the depuncturing pattern depending on the number of lost packets and depending on their position in the code word.

17. The device according to any one of claims 14 to 16, wherein the signal processing unit receives an information signal (BFI) indicating that a packet has been lost.

18. Device according to one of claims 14 to 16, in which the signal processing unit monitors a time period during which a packet has to arrive, the signal processing device indicating a loss of the packet when the time period has expired without the packet arriving.

19. Device according to one of claims 11 to 18 with a radio network (122) to transmit the receiving code word to a mobile receiver (124, 128), which comprises the signal processing unit.

20. The apparatus according to claim 19 with a coding unit (120) which re-codes the received code word before transmission to the mobile receiver.

21. A method for generating a code word, which contains code units which are generated based on information units by a redundancy code, for transmission via a lossy connection network, comprising the following steps:

(a) receiving the information units;

(b) encoding each information unit to generate a plurality of code units for each information unit; and (c) arranging the generated code units in the code word such that a loss of one or more code units in the code word leads to a code unit sequence which corresponds to a puncturing pattern applicable to the original code word, which is also known to a recipient of the code word ,

22. The method of claim 21, wherein

generating a plurality of packets containing code units, which are generated based on information units by a redundancy code, for transmission via a packet-oriented connection network (114),

in step (b) each of the information units is coded using a plurality of generators, the information unit being applied to each of the generators, and each generator generating a code unit for an adjacent information unit, and

in step (c) the code units are assigned to a plurality of packets, code units generated by the same generators being assigned to the same packets; and

the packets generated are made available for transmission.

23. Device for generating a code word, which contains code units which are generated based on information units by a redundancy code, for transmission via a lossy connection network, with:

an input for receiving the information units; an encoder (108) for encoding each information unit to generate a plurality of code units for each information unit; and

a processing device (110, 112) for arranging the generated code units in the code word such that loss of one or more code units in the code word leads to a code unit sequence which corresponds to a puncturing pattern applicable to the original code word, which also corresponds to a receiver of the Codeword is known.

24. The apparatus of claim 23 for generating a plurality of packets containing code units which are based on information units generated by a redundancy code for transmission over a packet-oriented connection network (114),

wherein the encoder (108) has a plurality of generators to which the information unit is applied, each of the generators generating a code unit for the applied information unit; and

wherein the processing means (110, 112), which receives from the encoder for each information unit connected to the encoder the assigned code units and assigns a plurality of packets, the processing means code units generated by the same generator in the encoder to the same packet assigns, and which provides the complete packets for transmission.

25. Transmission system to transmit information units with

a transmitter (102, 110) which receives the information units and a code word with a plurality from the information units by means of a redundancy code derived from code units, the transmitter comprising a device according to any one of claims 21 to 24;

a transmission medium (114) that receives and transmits the code words generated by the transmitter (102, 110); and

a receiver (116; 124, 128) which receives the code words transmitted from the transmission medium (114), the receiver comprising a device according to one of claims 11 to 20.

26. Transmission system according to claim 23, in which the transmission medium is a packet-oriented connection network (114).

27. The transmission system according to claim 24, wherein the transmission system further comprises a radio network (122), a base station (116) being arranged between the packet-oriented connection network (114) and the radio network (122), and the receiver being mobile is.