DE10345438B4 - Method and device for decoding packet-oriented data transmission networks transmitted coded data packets and method and apparatus for encoding and decoding of packet-oriented data transmission networks to be transmitted data packets - Google Patents

Abstract

Method for decoding coded data packets transmitted via packet-oriented data transmission networks with a plurality of data elements,
in which the coded data packets are subjected to error detection,
in which an error correction is carried out for the data packets identified as faulty, and
in which, for error correction, the data elements identified as being correctly transmitted are used for the data packets recognized as being defective
in which the decoder writes the data elements of each data packet into a matrix and reads them out orthogonally from the matrix for decoding.

Description

The The invention relates to a method and apparatus for decoding transmitted by packet-oriented data transmission networks encoded data packets and a method and apparatus for encoding and decoding over packet-oriented data transmission networks to be transferred Data packets.

For transfers of data over a network, e.g. above A computer network or a telephone network often becomes so-called packet-oriented data transfer applied. For this purpose, the data to be sent in the network in small Units, called packages, split. These packages contain in addition to the actual data content also information to correct them to the corresponding recipient to convey. Should now a large file on a packet-oriented network must first be divided into small packets, afterwards at the recipient to be reassembled.

at the transmission comes from data packets within data networks between sender and receiver it to transmission errors, causing problems in ensuring a desired transmission quality of the data leads. In other words to lead faulty transmission channels, in particular in mobile radio channels as used for example in mobile phones, to a deterioration the transmission quality.

Therefore have to Arrangements are made to prevent the occurrence of errors adulteration a signal, i. the occurrence of faulty data packets, if possible unlikely to make and mistakes that have occurred with as possible high probability to discover and, if necessary, corrective action to be able to initiate. This can only happen through the involvement of sender and receiver and requires appropriate protective measures. These protective measures, which are generally referred to as error protection however, expensive transmission capacities.

The the first and most important problem in the context of error protection is this Problem of reliable detection from mistakes. Only if errors are detected with high probability can, can also suitable measures be made to correct the error already occurred. A detection of transmission errors is only possible if the transferred Data has some redundancy, i. if additional Transfer information which, in the sense of data transmission, is "superfluous" information, but which provides information about it gives, whether a falsification the message is present, i. by means of which a faulty transmission can be recognized. A well-known method of such redundancy in the transmission to provide data is the provision of a so-called parity bit which to save ASCII or EBCDIC characters. Without Redundancy is a recognition of errors not possible in principle.

It There are various methods for detecting errors. They differ essentially only in the number of added "protection bits" and the safety, with the simple and complex disorders the original one Message and the protection bits are detected. A 100% guarantee for the recognition however, mistakes can never be made. Security is always relative. Furthermore, certainly only the presence of a Error be specified. Conversely, only with a certain, by the method certain probability, to be claimed that there is no error.

To the simplest method for error detection counts, as already mentioned, the introduction of parity bits. Here, each is to be transmitted Character with a check bit provided that the parity, i.e. indicates the number of 0 or 1 bits. In a blockwise transmission In addition to the cross, a so-called longitudinal parity can be checked.

For each m useful bits are transmitted k protection bits, so that a total of n = m + k bits over the Transfer line will go. It leaves prove mathematically that the parity bits do not have high protection offer, i. that security, with the errors that occurred too be detected, compared to other methods relatively low is. Methods Providing Improved Error Detection For example, the so-called cyclic redundancy codes (cyclical redundancy check = CRC), which is essentially a division by correspond to a polynomial with residual value formation.

• 1. Rückanforderung defekter, d.h. fehlerhaft übertragener, Pakete. Dies wird auch als „Acknowledge Mode" (ACK) bezeichnet. Dieser Prozess ist nur in Szenarien mit Rückkanal möglich, d.h. wenn zwischen einem Sender und einem Empfänger eine Zweiweg-Verbindung (Duplex-Verbindung) besteht. Bei Verwendung dieses Prozesses werden mittels eines Fehlererkennungsalgorithmus fehlerhafte übertragene Pakete erkannt. Die Fehlererkennung wird hierbei üblicherweise mittels eines zyklischen Redundanzcodes (CRC) durchgeführt. Die als fehlerhaft erkannten Pakete werden verworfen und der Empfänger fordert die fehlerhaft übertragenen Pakete erneut an. Nachteil bei diesem Verfahren ist jedoch, dass das erneute Versenden von angeforderten Paketen erhebliche Übertragungskapazität benötigt. Zusätzlich, wenn auch im geringen Masse, führt die erneute Übertragung einzelner Pakete zu einer Verzögerung der Übertragung der gewünschten Information, d.h. der gewünschten Daten.
• 2. Vorwärtsfehlerschutz, welcher auch als „Unacknowledge Mode" (UCK) bezeichnet wird. Beim Verwenden dieser Art von Fehlerschutz ist kein Rückkanal erforderlich. Damit ist dieser Prozess auch in Szenarien möglich, in denen kein Rückkanal vorhanden ist. Beim Vorwärtsfehlerschutz wird versucht, fehlerhafte Datenpakete mittels zusätzlich übertragener Korrekturinformationen, welche auch als Korrekturbytes bezeichnet werden, zu rekonstruieren. Hierbei ist zu beachten, dass jeder verwendete Vorwärtsfehlerschutz nur eine gewisse Korrekturfähigkeit aufweist, welche primär durch die Anzahl der mit übertragenen Korrekturbytes bestimmt wird. Bei der Vorwärtskorrektur verteilt ein so genannter „Interleaver" aufgetretene Paketverluste, d.h. fehlerhaft übertragene Datenpakete, auf einzelne Byte-Auslöschungen, die so genannten „Byte Erasures", welche korrigiert werden, wenn die Korrekturfähigkeit nicht überschritten wird. Bei der Auslöschungskorrektur werden somit Fehler eines als fehlerhaft erkannten Datenpakets auf einzelne Bytes verteilt, diese einzelnen Bytes werden dann ausgelöscht, wodurch die fehlerhaften Datenpakete korrigiert werden, wenn die Anzahl der Byte-Auslöschungen die Korrekturfähigkeit des verwendeten Codes nicht überschreitet.

For reliable data transmission, it is often not sufficient to be able to detect transmission errors; it is often necessary to be able to correct the data bytes which have transmission errors. For the correction of transmission errors, basically two different processes are known.

• 1. Request back defective, ie incorrectly transmitted, packages. This is also called "Acknowledge Mode" (ACK). This process is only possible in scenarios with return channel, ie if a two-way connection (duplex connection) exists between a sender and a receiver. Using this process, erroneous transmitted packets are detected by means of an error detection algorithm. The error detection is usually carried out by means of a cyclic redundancy code (CRC). The packets identified as faulty are discarded and the recipient requests the packets transmitted incorrectly again. Disadvantage of this method, however, is that the resending of requested packets requires significant transmission capacity. In addition, albeit to a lesser extent, the retransmission of individual packets results in a delay in the transmission of the desired information, ie the desired data.
• 2. Forward error protection, also referred to as "Unacknowledge Mode" (UCK) When using this type of error protection, no return channel is required, so this process is possible even in scenarios where no return channel is present It should be noted that each forward error protection used has only a certain corrective capability, which is primarily determined by the number of correction bytes transmitted with "Interleaver" packet losses, ie incorrectly transmitted data packets, to individual byte cancellations, the so-called "byte erasures", which are corrected if the correction capability is not exceeded s erroneously detected data packets are distributed to individual bytes, these individual bytes are then canceled, whereby the erroneous data packets are corrected if the number of byte cancellations does not exceed the correcting capability of the code used.

The forward error correction has the advantage that they are also performed on simplex channels can and that is a time consuming retransmission of data packets is avoided. But she has the disadvantage that always with high redundancy must be sent and not only in the Error case, a repetition takes place.

In the from the DE 100 34 977 A1 Known methods decompose defective data packets into sections and put them together to form combined new data packets in the hope that errors have occurred in the disjointed data packets in mutually disjoint parts, so that the resulting new data packets are error-free by combining the disjoint parts.

Of the The invention is based on the problem of transmission reliability in packet-oriented Increase networks, without that an increased transmission capacity additional Correction bytes or correction bits is necessary.

The Problem is solved by the method and apparatus for decoding from above packet-switched data transmission networks encoded data packets and by the method and the device for Encoding and decoding over packet-oriented data transmission networks to be transferred Data packets with the features of the independent claims solved.

at a method for decoding transmitted over packet-switched data transmission networks coded data packets having a plurality of data elements subjected the coded data packets to error detection. Further is for the erroneously recognized data packets carried out an error correction and for error correction, the data elements identified as being correctly transmitted are used the data packets recognized as faulty are used.

A Apparatus for decoding packet-switched data transmission networks coded data packets with a plurality of data elements, points a decoder arranged to code the encoded ones Data packets subjected to an error detection, for which recognized as faulty Data packets performs an error correction and error correction which are transmitted as correct recognized data elements of recognized as defective data packets uses.

In a method for coding and decoding data packets to be transmitted via packet-oriented data transmission networks, a plurality of data elements are transmitted by means of a Encoder coded to data packets, wel che data packets are transmitted over a packet-oriented data network. Furthermore, the coded data packets are subjected to an error detection with a plurality of data elements, an error correction is carried out for the data packets identified as being erroneous, whereby the data elements of the data packets identified as defective are used for error correction.

A Device for coding, transmission and decoding over packet-oriented data transmission networks to be transferred Data packets include an encoder and a decoder. Of the Encoder is arranged to have a plurality of data elements before transferring coded to data packets. The decoder is set up in such a way that it contains the coded data packets with a plurality of data elements undergoes a fault detection for which detected as faulty Data packets performs an error correction and error correction which are transmitted as correct recognized data elements of recognized as defective data packets uses.

The inventive method for decoding based on that even with a faulty data packet only a small number of data elements, which for example Data bytes or data bits, of the erroneous data packet are faulty, whereas the overwhelming Number of data elements correctly transmitted in incorrectly transmitted data packets become. By means of a simple algorithm it is possible the evaluate correct data elements of a faulty data packet. By the evaluation and use of the correct data elements of a faulty transmitted Data packets it is possible to increase the error protection, without doing the number mitübertrager Correction elements, i. Redundancy bytes or redundancy bits. The inventive method can be used in both error correction scenarios, i. Return request for defective packages and forward error protection, be used.

in the Process of return request Defective packets can by means of the method according to the invention, the number of newly requested packets are lowered. In the process of forward error correction can by means of the method according to the invention the number of correctable data packets are increased.

preferred Further developments of the invention will become apparent from the dependent claims. The further embodiments of the invention, which by way of the method for decoding packet-switched data transmission networks encoded data packets are also applicable to the device for decoding packet-switched data transmission networks coded data packets and for the method and apparatus for encoding and decoding over packet-oriented Data transmission networks to be transferred Data packets.

Preferably the error detection is performed in a unit of a communication transport layer.

The Error correction can be in a session layer or unit a presentation layer or a processing layer.

Under the layers are computer network layers, which can be found on layers in a transmission reference model, e.g. the OSI reference model or the TCP / IP reference model, Respectively. A lower layer than the transport layer corresponds for example the physical layer, the Link layer or the network layer in the OSI reference model or in the TCP / IP reference model. One layer higher than the transport layer is in the OSI reference model the session layer, the presentation layer, and the processing layer and in the TCP / IP reference model, the processing layer.

The Data elements can Be data bytes or data bits.

Especially Preferably, the decoder writes the data elements of each data packet column by column into a matrix and reads it line by line for decoding out of the matrix.

alternative For example, the decoder can divide the data elements of each data packet line by line write to a matrix and read out in columns for decoding. Illustratively, this means that the entire matrix is tilted, i. the columns and lines are reversed.

The fact that the decoder writes and reads the data to be transmitted in a different order in a matrix, ie either line by line and read column by column or vice versa, packet errors are converted into data element errors, which can be easily corrected in the following. In In the case of data element errors, the log-in is understood to mean, for example, byte errors or bit errors.

In A further development is the error correction of individual data packets carried out, if a redundancy by redundancy elements for performing a extinction correction the individual data packets is insufficient.

hereby Is it possible, that as long as an erasure correction (Erasure Correction) is feasible, because there is sufficient redundancy, first the cancellation correction carried out is before an error correction according to the invention carried out becomes. this leads to for improved error protection and improved correction the mistake. The number of redundancy elements can be per line or be different for each column.

following after the error correction according to the invention Individual data packets may require an additional cancellation correction be made.

Preferably will be after erasure correction recursively additional Bug fixes and erasure corrections performed.

That a recursive error correction is performed. By successful error correction individual data elements it is possible by means of the test information mark each column / row whole columns / rows as error-free. At a subsequent recursion step, i. to go through again error correction, these fall as error-free columns / rows path. This can be for some lines / columns from the correction mode, i. the error correction according to the invention, in the more reliable Erasuremodus, that is, the erasure correction, passed become. That it can additional Errors are corrected.

Especially The method is preferred for decoding with a soft decoding method combined.

Becomes the inventive method for decoding still combined with a soft decoding method, so will the error correction capability of the procedure additionally increased. Especially in the selection of potential to be corrected erroneous data elements can Here, those data elements are selected first, which is the lowest reliability or in other words the highest Probability of being faulty. This is possible in the transmission about a gain of 1 dB to 3 dB erziehlen [1].

In In a further development, the encoder writes the information to be transmitted Add data elements line by line to a matrix and read them for transfer as data packets from the matrix in columns, and the decoder writes the data elements of each data packet are split into a matrix and reads them out line by line for decoding from the matrix.

alternative the encoder can transmit the ones to be transmitted Write data elements column by column into a matrix and transfer them read as data packets line by line from the matrix and the decoder the data elements of each data packet line by line into a matrix write and read by column for decoding.

Thereby that the encoder is the one to transmit Writes and reads data in different order into a matrix, i.e. either line by line and read out column by column or conversely, packet errors are converted to data element errors, which can be corrected more easily below. This is special then the case, if the decoder this in exactly the reverse order performs, i. if the coder writes line by line in the matrix and column by column from the matrix, the decoder writes in columns the matrix and reads out line by line and vice versa.

Preferably the coder sets a different one per line or per column Number of redundancy elements ready.

Under Redundancy elements in the application, for example, redundancy bytes or redundancy bits, depending on whether from data bytes or of data bits is assumed. The redundancy elements are, if the encoder line by line in the matrix preferably writes each Row, whereas the redundancy elements when the encoder column by column in the matrix preferably writes per column provided become.

through the coding of redundancy elements, it is possible for the decoder incorrectly transmitted Corrected data elements of a data packet. This leads to a increase the transmission security. The Redundancy elements are used when the interleaver is written line by line can be added to the end of each line in the matrix and in addition to the end of each Column a check information added by which then at the receiver each column is checked for its defectiveness can be examined. In column-wise writing, the Redundancy elements are added to the end of each column and in addition are sent to the end of each Line a check information added, by which then at the receiver every line can be examined for their defectiveness.

The Number of redundancy elements can be individually per line or per column be determined. This makes it possible a uniform or uneven error protection to achieve. When using a different number of Redundancy elements it is possible the redundancy for particularly sensitive data elements within the data package and increase so the transmission security increase while it is possible at the same time is the number of redundancy elements for less sensitive data elements within of the data packet, without thereby the transmission security is severely impaired. Thus, an uneven error protection possible.

In summary the invention can be seen in that in the transmission data in packet-oriented networks increases the transmission security and the error rate is reduced by using a new method of error correction becomes. In this procedure, data packets recognized as erroneous not discarded but to an FEC decoder (forward error correction decoder, Forward error correction decoder) forwarded. It should be noted that data packets with the correct IP header Checksum but erroneous UDP checksum to the FEC decoder which, for example, in Network Adaptation Layer (NAL) is implemented. By means of a following in the detailed Description of an embodiment the algorithm described more precisely, the FEC decoder evaluates the correct bytes of a data packet identified as faulty. This will make a higher Error protection achieved.

that is, that erroneous data packets are not deleted, but in one higher Layer are transmitted as the transport layer on which then the inventive method for correction of transmission errors is applied. This can be, for example, the application layer. This is where the CRC check becomes no longer carried out in the transport layers as in the prior art, but only at a higher level Layer, preferably the application layer. This makes it possible the correctly transmitted Use data elements of a faulty data packet for error correction.

The inventive method can for a scenario with one available standing return channel be applied. In this scenario, it comes through the application the method according to the invention to a smaller number of re-requested data packets, because some of the data packets identified as faulty are corrected can be. Also for a scenario without one available standing return channel, For example, in television broadcasts, the inventive method be used. This is by the inventive method for error correction the number of errors reduced.

A preferred implementation of the inventive method for error correction is that for each Line of Data Bytes Three Different Types of Correcting Techniques of errors successively in a block interleaver, i. a decoder, be applied.

When first step is the conventional extinction correction uses. However, this is only applicable as long as the number of defective bytes of a packet not the erasure correction capability of the code, i. the transmitted Data packets, exceeds.

exceeds the number of defective bytes, the erasure correction capability, but is less than the error correction capability of the code, so will the inventive method used for error correction.

If the number of erroneous bytes exceeds the error correction capability of the code, the number of potentially defective bytes is reduced to the maximum erasure-correctable number by selecting a subset of the potentially erroneous bytes. Subsequently, the cancellation correction is performed and then the resulting bytes are verified by checksums. If the checksums indicate that the erase correction did not clean up the errors within the data packet, another subset is selected from the potentially erroneous bytes, and subsequently the erase correction is performed, followed by verification of the checksums. Selecting ver various subsets of potentially erroneous bytes may be performed until all the information contained in the redundancy of the code has been used to correct as many erroneous bytes as possible.

For the cancellation correction and the error correction according to the invention a so-called Reed-Solomon decoder can be used, which uses a so-called Reed-Solomon code. The Reed-Solomon Code is an error correcting method used in mobile communication in CDPD, in wired communication in ADSL and ATM, but also used for data storage on CDs and DVDs. The Reed-Solomon code is a forward error correction technique (FEC), which allows the correction of erroneously transmitted data blocks. The FEC data is encoded in a code word in a data block User data appended and transmit with them together. The codeword consists of the payload and a number of Reed-Solomon check bytes as well one or more DMT symbols. The mathematical description is similar to the of the CRC method using a previously formed Reed-Solomon check digit for the Polynomial Division is used.

Instead of a bytewise decoding, it is also possible according to the invention bitwise decoding, i. Error correction, make.

The Method of error correction can be accelerated in a simple manner become. The elaborate Part, i. the computationally expensive Part, is the erasure correction with reduced number of bytes. This part can be accelerated by not all combinations of subsets of potentially defective Bytes are selected but the selection changed Subsets is aborted after a predetermined period of time. Thereby will be the time for which the cancellation correction is used, limited, and the method of error correction accelerated. Here, however, the correction ability limited.

in the inventive method it is also possible so-called soft bytes, i. Bytes which have a slightly weakened level have to be recycled. This again increases the ability of the procedure to correct errors.

One embodiment The invention is illustrated in the figures and will be discussed below explained in more detail.

It demonstrate:

1 a schematic matrix for explaining an embodiment of the invention, in which matrix an encoder can write to be transmitted data packets;

2 a schematic frequency distribution from a simulation, which indicates the frequency distribution of byte errors per packet length;

3 a histogram in which frequency distributions of erroneous bytes per packet are entered;

4 Histograms in which the frequency distribution of erroneous bytes per packet for four different transmission conditions are entered;

5 how many additional lines of data bytes can be corrected when using an error correction method according to the invention;

6 a schematic representation of a simplified layer reference model with an explanation of the inventive idea; and

7 a schematic representation of a message transmission system.

Introducing becomes with reference to 7 a schematic message transmission system, ie a system for transmitting data, which has an encoder and a decoder and on which the inventive method for encoding and decoding can be performed.

In a message transmission system 700 is from a source 701 For example, a computer, a video camera, a landline telephone or mobile phone (not shown), a source encoder 703 , Source information 702 fed. The source encoder 703 generates a data stream from this 704 which comprises a plurality of payload data bytes.

Under Source coding is generally to be understood as a coding in the data to be transferred are compressed, i. superfluous redundancy in the to be transferred Data is eliminated.

The generated data stream 704 becomes an encoder 705 , also referred to as channel coder, supplied, coded by this and the coded data stream 706 becomes a modulator 707 supplied, in which the coded data stream 706 a carrier signal of predeterminable frequency is modulated and as a modulated signal 708 over a physical channel 709 , For example, via a radio link or via a telephone line, a receiver, a demodulator 710 has supplied.

For the purposes of this invention, coding or channel coding is to be understood as a method or a method by which a data stream is encoded by adding redundancy data, also referred to as redundancy elements in the application, so that from the source 701 to a sink 718 the data stream to be transmitted can be transmitted with the least possible number of errors. Thus, as part of the encoding of the encoder the actual user data controlled redundancy added to the side, so that when transmitting over the physical channel 709 resulting errors on the receiving side can be detected and corrected. The operation of the encoder will be described in more detail below.

In the physical channel 709 subject to the modulated signal 708 disorders 711 , causing a disturbed modulated signal 712 which is the demodulator 710 is supplied. The demodulator 710 generates a demodulated data stream from this 713 which is a decoder 714 , also called channel decoder, is supplied. The decoder 714 generated from the demodulated data stream 713 a decoded data stream 715 which is a source decoder 716 is supplied. The source decoder 716 generates a result data stream from the decoded data stream 717 , which is fed to a sink.

Based on 6 is schematically a layer reference model 600 , in the illustrated case a TCP / IP reference model, shown. In 6 is the third layer 601 ie, represents the network layer of the reference model. This third layer uses the Internet Protocol (IP). Further, the fourth layer 602 representing the transport layer of the reference model. Within the fourth layer, both the Transmission Control Protocol 603 (TCP) application as well as the User Datagram protocol 604 (UDP) and the real-time transport protocol 605 (RTP) application. Furthermore, in the 6 another fifth shift 606 representing the application layer, which in the OSI reference model is the seventh layer. Within the application layer 606 The various applications or applications can be executed such as the application H.323.

Based on 6 is also sketched the inventive idea. With the arrow in 6 it is indicated that IP packets (data packets) with correct IP header checksum but with a faulty UDP checksum are transmitted to the layer of the "Network Adaptation Layer (NAL)", ie the fifth layer (application layer), even if the UDP indicates checksum, The error correction method according to the invention, which will be described in more detail below, is then carried out within the NAL layer.

In 1 a schematic matrix 100 explains in which a coder can write data packets.

The illustrated matrix 100 consists of eleven lines 101 . 102 . 103 . 104 . 105 . 106 . 107 . 108 . 109 . 110 and 111 and 20 columns 112 . 113 . 114 . 115 . 116 . 117 . 118 . 119 . 120 . 121 . 122 . 123 . 124 . 125 . 126 . 127 . 128 . 129 . 130 and 131 , However, the number of rows and columns is not limited to these numbers. An encoder or interleaver writes the data bytes to be transferred into this matrix. Here he writes the data bytes line by line in the matrix. At the end of each line, it is possible for the encoder to insert redundancy elements, ie redundancy bytes. In the first three lines 101 . 102 and 103 be in the matrix of 1 12 data bytes and 8 redundancy bytes entered. In the fourth and fifth line 104 . 105 14 data bytes and six redundancy bytes are entered. In the sixth line 106 15 data bytes and five redundancy bytes are entered. In the seventh line 107 16 data bytes and four redundancy bytes are entered. In the eighth line 108 17 data bytes and three redundancy bytes are entered. In the ninth line 109 18 data bytes and two redundancy bytes are entered. In the tenth line 110 19 data bytes and one redundancy byte are entered. In the eleventh line 111 20 data bytes and no redundancy bytes are entered. The different number of redundancy bytes per line causes the individual lines have a different redundancy and thus are provided with a different error protection. In this way, important data bytes can be reconstructed better, more accurately and more easily in the event of a faulty transmission, as less important data bytes, which in the example are entered in one of the last lines, which have a lower number of redundancy bytes. It is thus realized an uneven error protection.

To transfer the data bytes are compiled into data packets after the line-by-line writing. To compile the data packets, the data bytes become column by column read. additionally becomes transferring still to every package check information added, which serves on the receiver side to recognize a faulty data packet.

In the in 1 Example shown are the sixth data packet, which through the sixth column 117 is formed, the ninth data packet, which through the ninth column 120 the tenth data packet, which is formed by the tenth column 121 the eleventh data packet, which is formed by the eleventh column 122 is formed and the 14th data packet, which through the 14th column 125 is formed, have been transmitted incorrectly and thus each have at least one erroneous byte. The erroneous data packets, ie the columns, are in the 1 characterized. In addition, in the 1 also the individual erroneous bytes within the faulty data packets marked with a cross.

Based on 1 can be easily seen that a byte error is generated by the different order of writing in the matrix (line by line) and reading from the matrix (columns) from a packet error, ie a data packet that was transmitted incorrectly. The check information of a data packet identifies in the 1 an entire column, ie an entire data packet, as faulty. However, since the bytes of data have been written into the matrix line by line and the redundancy bytes have been added line by line, a bad packet of data represents only a single byte of data within each line.

In general, for a correction of transmission errors follows:
For d ≤ t all errors of a line can be corrected by means of an erasure correction (erasure decoding).
For d> t, the FEC decoder can correct (t-1) / 2 bytes.

Or more generally:
For d> t, (8t-1) / 2 bits can be corrected by means of the FEC decoder.
Where: d = the number of bad packets and
t = the number of redundancy bytes.

following becomes an algorithm for a correction of errors as implemented in an FEC decoder can be explained.

• a) Der Algorithmus weist eine Zeilenschleife auf, d.h. eine Befehlsfolge, welche für alle Zeilen des Interleaver Blocks abgearbeitet wird.
• b) Jede Zeile wird zuerst auf potentiell defekte Bytes untersucht, d.h. es wird eine Fehlererkennung durchgeführt. Ferner wird entschieden, ob die Gesamtzahl der potentiell defekten Bytes die Erasure-Korrekturfähigkeit übersteigt.
• c) Wenn die Gesamtanzahl der potentiell defekten Bytes die Erasure-Korrekturfähigkeit nicht übersteigt wird eine Erasure Dekodierung, d.h. eine Auslöschungskorrektur durchgeführt. Nachdem eine Zeile korrigiert wurde, werden nochmals die CRC-Checks durchgeführt, d.h. die Zeile nach potentiell defekten Bytes untersucht und die Spalten des Interleaver Blocks, welche einen korrekten CRC-Check liefern als korrekt markiert. Nachfolgend wird zurück in die Zeilenschleife gesprungen und die nächste Zeile des Interleaver Blocks abgearbeitet, d.h. es wird zu b) zurückgesprungen.
• d) Wenn die Gesamtanzahl der potentiell defekten Bytes die Erasure-Korrekturfähigkeit übersteigt wird die Fehler-Dekodierung, d.h. die erfindungsgemäße Dekodierung oder er findungsgemäße Fehlerkorrektur, durchgeführt. Nach dieser Fehler-Dekodierung wird mittels der Checksummen überprüft ob die Fehlerkorrektur für die soeben korrigierte Zeile korrekt war. Hierbei wird überprüft ob die CRC korrekt sind, d.h. ob potentiell defekte Bytes in der Zeile vorhanden sind.
• e) Sind keine potentiell defekten Bytes in der Zeile vorhanden, so ist die Korrektur der Zeile abgeschlossen. Es werden nochmals die CRC-Checks durchgeführt, d.h. die Zeile nach potentiell defekten Bytes untersucht und die Spalten des Interleaver Blocks, welche einen korrekten CRC-Check liefern als korrekt markiert. Nachfolgend wird zurück in die Zeilenschleife gesprungen und die nächste Zeile des Interleaver Blocks abgearbeitet, d.h. es wird zu b) zurückgesprungen.
• f) Sind weiterhin potentiell defekten Bytes in der Zeile vorhanden, so wird eine Erasure Dekodierung mit reduzierter Bytezahl durchgeführt. Anschaulich bedeutet das, dass die Erasures, d.h. der Auslöschungen von potentiell fehlerbehafteten Bytes, willkürlich auf die maximal korrigierbare Anzahl reduziert wird. D.h. es werden willkürlich ausgewählte potentiell fehlerbehaftete Bytes ausgelöscht bis nur noch so viele Bytes übrigbleiben, wie maximal mittels der Erasure Dekodierung korrigiert werden können, wobei die entsprechende Anzahl von der Redundanz abhängt, welche in der Zeile vorhanden ist. Nachfolgend wird dekodiert und alle CRC Checksummen nochmals geprüft.
• g) Ergibt sich nun, dass die CRC Checksummen korrekt sind, d.h. keine potentiell defekten Bytes in der Zeile vorhanden sind, so ist die Korrektur der Zeile abgeschlossen. Es werden nochmals die CRC-Checks durchgeführt und die Spalten des Interleaver Blocks, welche einen korrekten CRC-Check liefern als korrekt markiert. Nachfolgend wird zurück in die Zeilenschleife gesprungen und die nächste Zeile des Interleaver Blocks abgearbeitet, d.h. es wird zu b) zurückgesprungen.
• h) Ergibt sich hingegen, dass die CRC Checksummen nicht korrekt sind, d.h. weiterhin potentiell defekte Bytes in der Zeile vorhanden sind, so wird praktisch zu Schritt f) zurückgesprungen, indem eine andere Teilmenge aus potentiell defekten Bytes ausgewählt und erneut eine Erasure Dekodierung versucht wird.

The algorithm is based on the data elements of an interleaver block, ie a matrix as shown in 1 is shown, applied several times. The application takes place until no further correction of data errors is possible, ie normally until all existing redundancy in the data for error correction is used.

• a) The algorithm has a row loop, ie a command sequence which is processed for all lines of the interleaver block.
• b) Each line is first examined for potentially defective bytes, ie an error detection is performed. It is further decided whether the total number of potentially defective bytes exceeds the erasure correction capability.
• c) If the total number of potentially defective bytes does not exceed the erasure correction capability, erasure decoding, ie, erasure correction, is performed. After a line has been corrected, the CRC checks are carried out again, ie the line is examined for potentially defective bytes and the columns of the interleaver block which provide a correct CRC check are marked as correct. Subsequently, the system jumps back to the line loop and executes the next line of the interleaver block, ie it jumps back to b).
• d) When the total number of potentially defective bytes exceeds the erasure correcting capability, the error decoding, ie decoding according to the invention or error correction according to the invention, is performed. After this error decoding is checked by means of the checksums whether the error chooser correction was correct for the line just corrected. In this case, it is checked whether the CRC are correct, ie whether there are potentially defective bytes in the line.
• e) If there are no potentially defective bytes in the line, the correction of the line is completed. The CRC checks are carried out again, ie the line is examined for potentially defective bytes and the columns of the interleaver block which supply a correct CRC check are marked as correct. Subsequently, the system jumps back to the line loop and executes the next line of the interleaver block, ie it jumps back to b).
• f) If there are still potentially defective bytes in the line, erasure decoding is performed with a reduced number of bytes. Illustratively, this means that the erasure, ie the deletion of potentially error-prone bytes, is arbitrarily reduced to the maximum correctable number. That is, randomly selected potentially erroneous bytes are erased until only as many bytes remain as can be corrected at most by erasure decoding, the corresponding number depending on the redundancy that exists in the line. Subsequently, decoding and checking all CRC checksums again.
• g) If it now turns out that the CRC checksums are correct, ie there are no potentially defective bytes in the line, the correction of the line is completed. The CRC checks are carried out again and the columns of the interleaver block, which deliver a correct CRC check, are marked as correct. Subsequently, the system jumps back to the line loop and executes the next line of the interleaver block, ie it jumps back to b).
• h) If, on the other hand, it results that the CRC checksums are not correct, ie there are still potentially defective bytes in the line, then in practice step f) is jumped back by selecting another subset of potentially defective bytes and attempting an erasure decoding again ,

The algorithm described above is given below in a pseudocode:
Loop over interleaver block, eg a matrix like in 1 shown:
Execution of the interleaver block several times until no further correction is possible

Based of the pseudocode is easily recognizable as the algorithm the correction of transmission errors according to one embodiment of the invention is implemented. The shown algorithm for reconstruction incorrectly transmitted Data packets are based on the correct bytes of the respective faulty ones Data packets on. It should also be noted that it is also possible Swap columns and rows in the algorithm. This is then necessary, when the interleaver writes the data bytes into the matrix column by column and to transfer read out line by line.

The following is based on the 2 to 5 , which were obtained by means of simulations, describes how he an inventive error correction method on the transmission quality, ie the number of errors after the error correction, existing effect. The simulations were based on a packet length of 20 bytes.

In 2 schematically the results of a simulation are shown, which indicates the frequency distribution of byte errors per packet length. Plotted is the frequency with which a certain number of byte errors / packet length occurs. It can be seen that 0.05 byte error / packet length, ie an error rate of one wrong byte per 20 byte packet length, occurs in about 5 ‰ of all cases. The frequency of occurrence of a particular ratio of byte error / packet length increases up to the ratio of 0.25 byte error / packet length and reaches a frequency of about 2.6% there. Thereafter, the frequency of occurrence of a certain byte error / packet length ratio drops to the 0.45 byte error / packet length ratio where it reaches a frequency of about 3 ‰.

In the 2 The simulation shown was carried out in the case of a mobile radio transmission, eg mobile telephone conversations, and shows that on average about 11% of the data packets have faulty data bytes, ie that only about every ninth data packet is transmitted incorrectly. The 11% is the sum of all in the 2 shown frequencies. The maximum frequency is a byte error / packet length ratio of 0.25, ie one byte out of four byte packet length, which means that each the fourth byte is faulty. However, here too, only a frequency of about 2.6% results. Thus, it is easy to see that most data packets are transmitted without errors and also that the overwhelming number of bytes are correctly transmitted in the errored data packets.

3 shows a histogram in which the frequency distribution of erroneous bytes per packet is entered. The absolute number of erroneous bytes on the X-axis is arbitrary here, since the number of bytes in error depends on the packet length used. Registered is the frequency distribution for faulty bytes, which could not be corrected despite the correction of errors. To get the data points of the top curve 340 a correction of the transmission errors has been carried out according to the prior art, in which the correct bytes of the erroneous data packets are not used for the correction of the errors. To get the data points of the lower curve 241 an error correction has been carried out according to the method according to the invention, in which the correct bytes of the erroneous data packets are used for the error correction. It can clearly be seen in the histogram that the error correction method according to the invention results in a significantly reduced number of erroneous bytes per data packet than the method according to the prior art. The curve 341 is shifted significantly to the left, which means that a larger number of initial errors can be corrected when the error correction method according to the invention is used.

4 shows four histograms ad, in which, as in 3 , Frequency distributions of erroneous bytes per packet is entered. The absolute number of bytes of errors on the X-axis is again arbitrary, since the number of bytes in error depends on the packet length used. Registered is the frequency distribution for faulty bytes, which could not be corrected despite the correction of errors. In the 4a and 4b Transmission conditions are selected which correspond to a transmission within a building, that is, for example, when a mobile phone is operated within a building. To determine the points of the upper curves 440 and 442 a correction of the transmission errors has been carried out according to the prior art, in which the correct bytes of the erroneous data packets are not used for the correction of the errors. To determine the points of the lower curves 441 and 443 an error correction has been carried out according to the method of the invention in which the correct bytes of the erroneous data packets are used for the error correction. It can be clearly seen in the histograms that the correction method according to the invention results in a significantly reduced number of erroneous bytes per data packet than the method according to the prior art. The curves 441 and 443 are shifted significantly to the left, which means that a larger number of initial errors can be corrected when the error correction method according to the invention is used.

The same applies to the in the 4c and 4d shown examples which correspond to a transmission situation within a vehicle, that is, for example, when a mobile phone is operated within a vehicle. Again, the upper curves correspond 444 and 446 a correction of the transmission errors according to the prior art, in which the correct bytes of the erroneous data packets are not used for the correction of the errors. Whereas, to determine the points of the lower curves 445 and 447 an error correction has been carried out according to the inventive method in which the correct bytes of the erroneous data packets are used for the error correction. In the histograms of 4c and 4d It can clearly be seen that the correction method according to the invention results in a significantly reduced number of erroneous bytes per data packet than the method according to the prior art. The curves 445 and 447 are shifted significantly to the left, which means that a larger number of initial errors can be corrected when the error correction method according to the invention is used.

In 5 FIG. 3 illustrates how many additional lines of data bytes can be corrected when an error correction method according to the invention is used. More specifically, it shows how many adjacent correct lines within an interleaver block, ie a block, as shown schematically as a matrix in 1 can be expected when an error correction method according to the invention is used. The probability of a certain number of consecutive error-free rows can be calculated in a simple way as follows: P (n) = (1 - p) n in which:

n

the number of each other subsequent is error-free lines;

p

the error probability a single line; and

P (n)

the probability for n together subsequent is error-free lines.

In the 5 The probabilities shown are based on those in 3 presented results. In the 5 lower curve 550 shows the probability of the occurrence of consecutive error-free lines. When the inventive error correction method is used without providing redundancy bytes, about three additional consecutive error-free lines are expected. The number of three results from the probability value of 50%, ie there is more than 50% improvement by three additional consecutive error-free lines. Thus, it can be seen that the error correction can be improved by means of the error correction method according to the invention, which uses the correct bytes of erroneous data packets, by also using the errored data packets for error correction. This achieves greater error protection and greater transmission reliability.

The upper curve 551 in 3 shows how the probabilities for successive error-free lines behave, if by using redundancy bytes one error per line can be corrected. In this case, about 20 consecutive good lines are expected. The number of 20 again results from the probability value of 50%, ie more than 50% result in 20 consecutive error-free lines. It can easily be seen from these curves that the error correction method according to the invention leads to a strong improvement of the correction capability and thus to an improvement of the error protection and an improvement of the transmission reliability.

Sum up the method according to the invention is based for decoding that even with a faulty data packet only a small number of data elements, which for example Data bytes or data bits, of the erroneous data packet are faulty, whereas the vast majority of data elements correctly transmitted in incorrectly transmitted data packets become. By means of a simple algorithm it is possible the evaluate correct data elements of a faulty data packet. By evaluating and using the correct data elements of a incorrectly transmitted data packets Is it possible to increase the error protection, without doing the number mitübertrager Correction elements, i. Redundancy bytes or redundancy bits. The inventive method can be used in both error correction scenarios, i. Return request for defective packages and forward error protection, be used.

• [1] Performance of Soft Decision Digital Matched Filter in Direct-Sequence Spread-Spectrum Communication Systems, N. Kataoka, T. Kojima, M. Miyake and T. Fujino, IEICE Transactions Vol. E 74, Nr. 5, (1991) 1115-1122

This document cites the following document:

• [1] Performance of Soft Decision Digital Matched Filter in Direct Sequence Spread Spectrum Communication Systems, N. Kataoka, T. Kojima, M. Miyake and T. Fujino, IEICE Transactions Vol. E 74, No. 5, (1991) 1115-1122

Claims (23)

Method for decoding transmitted over packet-oriented data transmission networks encoded data packets with a plurality of data elements, at the coded data packets are subjected to error detection, at for the erroneously recognized data packets are error corrected, and in which the error correction as the correct transfer recognized data elements of recognized as defective data packets be used where the decoder the data elements each Data packet writes into a matrix and orthogonal for decoding read from the matrix. Method according to claim 1, in which the error detection in a unit of a communication transport layer carried out becomes. Method according to claim 1 or 2, in which the error correction in a unit, a session layer or a presentation layer or an application layer is performed. Method according to one the claims 1 to 3, in which the data elements are data bytes. Method according to one the claims 1 to 3, in which the data elements are data bits. Method according to one the claims 1 to 5, in which the error correction of individual data packets is performed, if a redundancy by redundancy elements for performing a extinction correction the individual data packets is insufficient. Method according to claim 1, in which after the error correction of individual data packets an erasure correction is made. Method according to claim 7, in which after the erasure correction recursively additional Bug fixes and erasure corrections be made. Method according to one the claims 1 to 8, in addition to a soft decoding method is provided. Method according to one the claims 1 to 9, in which the error correction of the decoder in the network Adaptation Layer takes place. Method for coding and decoding via packet-oriented Data transmission networks to be transferred Data packets in which a plurality of data elements before transmitting be encoded into data packets by means of an encoder, at the coded data packets over transmit a packet-oriented data network become, in which the coded data packets with a plurality be subjected to error detection of data elements, at for the erroneously recognized data packets are error corrected, and in which the error correction as the correct transfer recognized data elements of recognized as defective data packets be used, where the encoder is the data elements of each Data packet writes into a matrix and orthogonal for decoding read from the matrix. Method according to claim 11, in which the error detection of a unit of a communication transport layer carried out becomes. Method according to claim 11 or 12, in which the error correction in a unit, a Session layer or a presentation layer or an application layer carried out becomes. Method according to one the claims 12 to 13, in which the data elements are data bytes. Method according to one the claims 12 to 13, in which the data elements are data bits. Method according to one the claims 12 to 15, in which by the encoder per row or per column provided number of redundancy elements is different. Method according to one the claims 12 to 16, in which the error correction of individual data packets is performed, if a redundancy by redundancy elements for performing a extinction correction the individual data packets is insufficient. Method according to claim 17, in which after the error correction of individual data packets a extinction correction is made. Method according to claim 18, in which after erasure correction recursive additional bug fixes and cancellation corrections be made. Method according to one the claims 12 to 19, in addition to a soft decoding method is provided. Method according to one the claims 12-20, in which the error correction of the decoder in the network Adaptation Layer takes place. Device for decoding coded data packets transmitted by packet-oriented data transmission networks with a plurality of data elements, with a decoder which is set up in such a manner that the following method steps can be carried out: an error correction is carried out for the data packets identified as faulty, an error correction is carried out for the data packets recognized as faulty, and the data elements identified as faulty are used for error correction as data packets identified as being incorrectly the data elements of each data packet are written into a matrix and orthogonally read from the matrix for decoding. Device for coding, transmitting and decoding over packet-oriented Data transmission networks to be transferred Data packets with an encoder and a decoder, in which the encoder is arranged to have a plurality of data elements before transferring encoded into data packets, the data elements of each data packet are written in a matrix and coded orthogonally from the Matrix read out and in which the decoder is set up in this way is that he the coded data packets with a plurality of Subjecting data items to error detection, for as erroneously recognized data packets performs an error correction, and to Error correction which is transmitted as correct recognized data elements of recognized as defective data packets uses.