WO2002097603A1

WO2002097603A1 - Method for analysing a digital audio flow and device for receiving said flow

Info

Publication number: WO2002097603A1
Application number: PCT/FR2002/001827
Authority: WO
Inventors: Edouard Ritz
Original assignee: Thomson Licensing S.A.
Priority date: 2001-06-01
Filing date: 2002-05-31
Publication date: 2002-12-05

Abstract

The invention relates to a method for analysing a digital audio flow comprising a stage in which part of a digital flow is compared to at least one specific signature of a digital audio format. The invention also relates to a device for receiving one such flow.

Description

Method for analyzing a digital audio stream and device for receiving such a stream

The present invention relates to a method for analyzing a digital audio stream as well as a device for receiving such a stream.

A digital audio stream is a stream of information encoded in digital form (that is to say basically in the form of bits) and which represents an audio signal, that is to say a signal intended to be reproduced as an acoustic signal.

Such a digital audio stream is for example received by a digital decoder as the soundtrack of an audio-video program, for example a television program. Conventionally, such a digital audio stream is coded according to a single predetermined standard, for example MPEG2, known to the digital decoder (in general, moreover, because the digital decoder is supplied by the transmitter of the digital audio-video source which encoded the stream).

More recently, other standards (or formats) for digital audio coding have developed, such as, for example, AC-3 and WAVE. The format in which digital audio information is encoded (i.e. the format of the audio file for microcomputing) is available as an indicator, such as the file extension for microcomputing . Thus, the central unit of a microcomputer has access to the format used by reading the extension of the audio file, for example '.wav' for the WAVE format. Similarly, information relating to the audio format used is available during the transmission of data by a digital interface of the IEC 61937 type, according to which the audio data is encapsulated with the addition of information relating to the format.

The inventor has realized that, on the contrary, a problem arises when it is desired to transmit digital audio data of variable coding format, which can therefore be raw data, in a digital stream. This is of course desirable in an advanced multimedia environment. On the receiver side (where decoding should take place) - for example a digital decoder, the problem arises in this way: it is a matter of being able to decode data from a digital audio stream when no information on the format is only available in the digital stream. To remedy this problem, the inventor proposes a method for analyzing a digital audio stream which comprises a step of comparing part of the digital stream with at least one specific signature of a digital audio format. According to other advantageous characteristics,

- The method includes the following additional step when said comparison is negative: comparison of a second part of the digital stream to a second specific signature of a second digital audio format; - said comparison step is repeated regularly in order to detect a dynamic change in format;

- the digital audio stream is part of an audio-video stream received from a broadcaster;

- said part of the digital stream is part of raw audio data.

There is also proposed a device for receiving a digital audio stream which comprises means for comparing part of the digital stream with at least one specific signature of a digital audio format. According to advantageous characteristics,

the comparison means have access to a plurality of signatures for comparison with respective parts of the digital stream;

- control means allow the regular implementation of the comparison means;

- an audio decoder is configured as a function of format information supplied by the comparison means;

- The device comprises means for receiving a continuous audio video stream. The following description will be made with reference to the accompanying drawings in which:

- Figure 1 shows a header of an audio frame in MPEG ES format;

- Figure 2 shows a header of a PES packet of an MPES PES stream;

- Figure 3 shows the structure of an audio frame in AC-3 format; - Figure 4 shows the structure of an audio stream in WAVE format;

- Figure 5 shows the flowchart of automatic format detection. An example of a method is described below, implemented for example in software form, making it possible to automatically detect the format of a broadcast digital audio stream (in the form of a continuous data stream, or of data packets which follow in time), and extract all the parameters (sampling frequency, bit rate, number of channels, type, selected channels, prologic mode, number of front, rear channels ...) allowing automatic configuration of a decoder multi-format audio to play the stream in question. The example is described below for the following audio formats:

MPEG 1 and 2 (layer I, II, III) Elementary Stream (ES) MPEG 1 and 2 (layer I, II, III) Packetized Elementary Stream (PES)

AC-3 Elementary Stream (ES) AC-3 Packetized Elementary Stream (PES) WAVE but can be easily extended to other digital audio formats.

The automatic detection of digital audio format described in this document is a technique that can intervene in a fairly wide application framework, since this method can be implemented and adapted to all types of data dissemination, regardless of the transport medium used. (satellite, cable and terrestrial broadcasting, packet transmission over computer networks, data flow - streaming - over the Internet, etc.). The reason is that this method works on any binary data stream (sequence of 0 and 1).

The context of use is as follows: A transmitter (digital broadcasting center, server, etc.) transmits or broadcasts a binary stream of audio data in a given format, unknown to the receiver (s). Each receiver is equipped with a multi-format and configurable audio stream decoding system (hardware or software). The receiver receives the data, must automatically detect the format of the audio stream from a list of supported formats, and extract the decoding parameters to configure the audio decoder, so that it can play the audio stream in question. The decoder must also be able to detect a dynamic change in audio format occurring during the broadcast. For example, an MPEG ES audio stream is broadcast. The receiver receives the data, analyzes it, detects that the format of the stream is MPEG ES, configures the decoder and starts playing the MPEG stream. Along the way, a format change occurs and an AC3 stream is broadcast while the receiver is still decoding the MPEG stream. On reception of the first AC3 data, the receiver detects the format discontinuity and performs the appropriate processing (it can report the error, ignore the AC3 data, reconfigure itself to play the new AC3 stream, etc.)

In order to implement the format detection method, it is proposed that the receiver knows a priori (that is to say before any reception) the syntax of the bitstream data stream (bitstream) of a certain number of audio formats , and therefore has at the outset a list of formats that it can identify. In the example, the list of supported formats is as follows:

MPEG 1 and 2 (layer I, II, III) Elementary Stream (ES) MPEG 1 and 2 (layer I, II, III) Packetized Elementary Stream (PES) AC-3 Elementary Stream (ES) AC-3 Packetized Elementary Stream ( PES)

WAVE

The detection problem therefore amounts to selecting from among the formats known to the receiver (cf. list above) that which corresponds to the data received. If the data received does not follow the syntax of the formats in this list, detection is impossible.

The actual detection method is based on the following principle:

For each of the formats supported by the receiver (see list above), a signature characteristic of the format is defined. This signature, extracted from the binary train syntax of each audio format, makes it possible to identify it in a unique and certain way. Once these signatures have been defined, the detection algorithm consists in analyzing the binary data stream received to iteratively search for each signature in a given order: we start by searching for the signature of the MPEG ES format, if we fail, we search for the signature of the following format and so on until all the formats in the list have been exhausted. Detection succeeds if a signature has been found. It fails if no signature was found after browsing the list of all known formats of the receiver. In which case, we know with certainty that the stream received does not belong to any of the known formats (but we do not know what its format is).

In the following, a possible signature is defined for each format.

MPEG ES format:

An audio frame of an MPEG elementary stream (ES) begins with a 32-bit header consisting of the fields shown in Figure 1.

The synchronization pattern (Syncword) is coded on 12 bits and the standard fixes its value at OxFFF (in hexadecimal), the standard guarantees that this pattern appears only in the header and nowhere else.

Furthermore, an MPEG header is invalid if at least one of the following fields takes a value reserved by the standard:

Layer = 00 Bitrate index = 11 1 1

Sampling frequency = 1 1

Emphasis = 10

The signature of an MPEG ES stream can therefore be defined by the following condition: If in the binary data stream received, there is a 4-byte word starting with OxFFF and whose bits:

- 14 and 15 (Layer) differ from '00'

- 17 to 22 (Bitrate index and Sampling frequency) differ from '11 1 1 1 1' - 31 and 32 (Emphasis) differ from '10' so the stream is in MPEG ES format.

In other words, the signature of an MPEG ES stream is equivalent to the detection in the bit stream of a valid MPEG header.

MPEG PES format The PES layer of the MPEG standard is a system layer used to transport audio-video data. It is an overlay of the ES layer (Elementary Stream). A PES packet begins with a PES header containing the fields shown in Figure 2.

The standard fixes the value of the Packet Start Code Prefix at 0x000001 and guarantees that this code cannot be encountered elsewhere in the flow than in the PES Header. This code indicates the start of a PES packet. The Stream Id field, coded on 1 byte, is used to identify the type of elementary stream (ES) contained in the PES packet. A value of Stream d equal to '110xxxxx' where x can be equal to 0 or 1 identifies an MPEG audio elementary stream 1 or 2. So any value of Streamjd between OxCO and OxDF (inclusive) identifies an MPEG audio stream.

The PES Header Length field is used to determine the length of the PES header, and therefore to determine where the start of the Payload of the PES packet is located. The Payload of an audio PES packet contains the data of the elementary stream and therefore begins with a header. Therefore, if in the data bitstream there are 3 consecutive bytes taking the hexadecimal value 0x000001 followed by 3 bits taking the binary value '1 10', then the stream is an MPEG PES audio stream. As a precaution, we add to this signature that of an MPEG ES stream, Le. it is verified that the first audio frame contained in the PES packet begins with a valid MPEG header. We thus define the signature of the MPEG PES audio format:

If in the received binary data stream, the binary sequence '000000000000000000000000001010' is found and the Payload of the current PES packet begins with a valid MPEG header (see above), then the stream is in MPEG PES format. AC3 ES flow:

The structure of an AC3 audio frame is shown in the diagram in Figure 3.

The AC-3 stream consists of successive synchronization windows in which each part represents information important for decompression and data recovery.

SI block: represents the synchronization information. BSI block: contains information on the type of coding and on the data structure. Block AB n: each block contains the audio data of the different channels. Each block consists of 256 sound samples.

Aux block: this part can contain additional information on the previous blocks; this information is used when emergency data is needed. CRC block: the error check makes it possible to check that the information is not incorrect. A CRC code is also found in the SI frame in order to reduce the number of false synchronization frames.

SI contains a 40-bit synchronization word used to indicate the start of the AC3 frame. This word is at the start of each frame. These 40 bits are divided into 4 variables: Syncword, Crc1, Fscod and Frmsizecod.

Syncword (16 bits) - synchronization word - is always 0x0B7 i.e. '0000 1011 0111 0111'. The bit order in the stream is left bit first. Crc1 (16 bits): error control code is used for 5/8 of the frame. The most significant bit first during transmission.

Fscod (2 bits): this variable indicates the frequency of the sample. The code '11' is reserved by the AC3 standard.

Frmsizecod (6 bits): this word determines the number of 16-bit words preceding the next frame

BSI contains information about the type of data carried in the flow. It is only from this data that it is possible to reconstruct the original samples. Less important information is also transported, such as language, time, type of service (dialogue, comment, music, etc.).

Bsid - Stream authentication (5 bits): in the current version of Dolby Digital coding, the value of Bsid is '01000' (= 8).

Bsmod - Data type (3 bits): these 3 bits associated with Acmod allow to indicate the type of service transported in the stream. Acmod - Audio coding mode (3 bits): this code is used to determine the number of voices used. It is possible to encode up to 6 channels. The six channels available are: central channel C, front left channel L, front right channel R, rear left channel SL, rear right channel SR, box LFE. The box is not counted in the number of channels but is activated by setting the Ifeon variable to 1.

Cmixlev - Sound level of the central channel (2 bits): these two bits are used to indicate the attenuation to be achieved on the central channel.

Surmixlev - Noise level of the rear channels (2 bits): these two bits indicate the attenuation to be achieved on the rear channels. Dsurmod - Dolby surround mode (2 bits): these two bits indicate the presence of a Dolby surround mode encoded in the left and right channels.

Lfeon - Low frequency channel (1 bit): This bit is set to 1 when the specific subwoofer channel is present in the stream.

The AC3 ES format is identified by its 16-bit synchronization word, which is always 0x0B77 in hexadecimal. If this pattern is identified in the received bit stream and if the information in the SI and BSI block that follows (the Syncword marks the start of the SI block.) Is valid, e. that no field takes the value reserved by the standard, then the audio stream received is in AC3 ES format. The signature of the AC3 ES format is therefore defined as follows:

Syncword '0x0 B77' found in the stream + valid SI and BSI block.

Furthermore, the decoding of the SI and BSI blocks makes it possible to extract the decoding parameters to configure the audio decoder.

AC3 PES format:

An AC3 PES stream begins with a Startcode 0x000001 and has a Streamid value set to OxBD (private_stream_1 in the standard

MPEG) by ATSC standard. To this condition, the signature of an AC3 ES stream, Le, is added for greater safety. we make sure that the Payload of the package

PES starts with Syncword 0x0B77 and that the SI and BSI blocks are valid.

If in the binary data stream there is a 4 byte word

(Startcode and Streamid) taking the hexadecimal value 0x000001 BD, and the Payload of the current PES packet begins with the Syncword 0x0B77 and the following information corresponds to valid fields of the SI and BSI blocks, then the audio stream received is in AC3 format SEP.

WAVE format

An audio stream in WAVE format has the structure shown in Figure 4. A WAVE file is characterized by the value of the following fields:

ChunklD = RIFF (0x52494646 in hexadecimal)

Format = WAVE (0x57415645 in hexadecimal)

SubchunklID = "fmt_" (0x666d7420 in hexadecimal) AudioFormat = 1

Subchunk2ID = data (0x64617461 in hexadecimal)

These values constitute the signature of a WAVE file. Therefore, if in the given bitstream we successively find: the word 0x52494646 at an address marked ADR the word 0x57415645 at the address ADR + 8 bytes the word 0x666d7420 at the address ADR + 12 bytes the half-word 0x0001 at the address ADR + 20 bytes the word 0x666d7420 at the address ADR + 36 bytes then the audio stream is in WAVE format.

Once the signatures of each audio format have been defined, automatic detection comes down to a simple search for signatures in a binary data stream. This search is carried out according to the flow chart represented in FIG. 5.

Claims

1. A method of analyzing a digital audio stream which comprises a step of comparing part of the digital stream with at least one specific signature of a digital audio format.

2. Method according to claim 1 which comprises the following additional step when said comparison is negative: - comparison of a second part of the digital stream to a second specific signature of a second digital audio format.

3. The method of claim 1 or 2, wherein said comparing step is repeated regularly to detect a dynamic change in format.

4. Method according to one of claims 1 to 3, wherein the digital audio stream is part of an audio-video stream received from a broadcaster.

5. Method according to one of claims 1 to 4, wherein said part of the digital stream is part of raw audio data.

6. Device for receiving a digital audio stream, characterized in that it comprises means for comparing a part of the digital stream with at least one specific signature of a digital audio format.

7. Device according to claim 6, in which the comparison means have access to a plurality of signatures for comparison with respective parts of the digital stream.

8. Device according to claim 6 or 7, wherein control means allow the regular implementation of the comparison means.

9. Device according to one of claims 6 to 8, in which an audio decoder is configured as a function of format information supplied by the comparison means.

10. Device according to one of claims 6 to 9, which comprises means for receiving a continuous audio-video stream.