EP0891617B1 - Signal coding and decoding system, particularly for a digital audio signal - Google Patents

Abstract

PCT No. PCT/FR97/00582 Sec. 371 Date Apr. 27, 1999 Sec. 102(e) Date Apr. 27, 1999 PCT Filed Apr. 2, 1997 PCT Pub. No. WO97/38417 PCT Pub. Date Oct. 16, 1997A coding system delivers a global data stream consisting of primary coded subband data streams from a primary subband coder bank, coding an input signal data stream, and secondary coded subband data streams from a secondary subband coder bank. The coding delay of the primary coder bank is smaller than that of the secondary coder bank. A filter bank receives the input signal data and generates signal streams in a plurality of subbands, which are coded by the respective coder of the primary subband coder bank, forming the primary streams. A bank of decoders receive and decode the respective coded primary subbank streams, which decoded subband signals are subtracted by a bank of subtractors from the corresponding original subband signals, which difference streams are input to the respective coder in a secondary subband coder bank. The secondary coder generates coded secondary subband data streams. A multiplexer interlaces the primary and the secondary coded subband data streams into a single global data stream.

Description

The present invention relates to a coding and decoding of a signal, in particular of a digital audio signal. These systems find application in low speed transmission of audio signals, with a coding / decoding delay constraint as low as possible imposed for example by the return of a voice control.

When transmitting digital signals, these are digitally coded in the transmitter then decoded, for their restitution, in a receiver. The present invention is concerned by the antinomy between, on the one hand, the search for a quality of transmission which generally entails for a fixed speed a delay of relatively long coding and decoding and, on the other hand, the delay of coding / decoding which, in some applications, must be short.

In the present description, the term delay coding / decoding the time between the entry of a sample into the encoder of the sample output corresponding to the decoder. For get rid of the particular implementation of the coding process and / or the structure of the circuits which allow this coding, will consider that the calculations made during these processes are infinitely fast both within the encoder and the decoder. Alone are therefore taken into account in the calculation of the coding / decoding delay parameters such as the duration of acquisition of signal frames digital, the delay imposed by a filter bank and / or the duration corresponding to a multiplexing of samples.

In the case of a transform coder, this delay will be greater than the duration of a coded frame added to the delay generated by the transform. In the case of a low delay coder of the LD-CELP type such as that described by JHChen and all in the article entitled "A low delay CELP coder for the CCITT 16kb / s speech coding standard" published in IEEE J. Salt. Areas Commun., Vol 10, pp 830-849, the delay is linked to the five samples constituting a basic frame. Note that a coding scheme has a delay expressed in number of samples. To deduce a time value, the sampling frequency at which the encoder is operated must be used, according to the relationship: time duration = delay in samples sampling frequency

As for the quality of coding, it is a parameter which is difficult to define, knowing that the final receiver, i.e. the ear of the auditor cannot give precise quantitative results. Through elsewhere, measurements such as the signal-to-noise ratio do not are irrelevant because they do not take into account the properties of psycho-acoustic masking of the auditory system. Techniques statistics such as those recommended by ITU-R-BS-1116 differentiate between different coding algorithms with regard to the quality of coding.

It should be noted, however, that an improvement in the signal to noise produced on all the frequencies of the sound signal allows to improve the perceived quality.

Coding systems for generic digital audio signals, that is to say without assumption on the mode of production of these signals, have so far taken as little constraint the time aspect of signal reconstruction. An exception is however illustrated by the process which is described by F. Rumseyi in an article entitled "Hearing both sides-stereo sound for TV in the UK" published in IEE rev, vol 36, No 5, pp173-176. However, in this process, the rates of compression achieved do not allow to compete with coders to classic transforms.

Among the algorithms that are standardized in ISO (ISO / IEC 13818-3), the minimum reconstruction times range from 18 ms for the simplest - and therefore least efficient - encoder at more than 100 ms for the most complex coder. Other non-coding methods standardized by ISO such as the so-called AC3 process described by C. Todd and all, such as the process known as ASPEC (Adaptive Spectral Perceptual Entropy Coding) described by K. Brandebug and all, or the method said ATRAC (Adaptive Transform Acoustic Coding) described by K. Tsutsui typically have coding / decoding delays of the order of one hundred milliseconds.

The effectiveness of coding systems is linked to the size of the filter banks which are generally used, taking into account long-term redundancies in the signals to be coded, at the optimal distribution of binary allocations over a period higher than the frame, etc. Taking these elements into account when coding time increases the delay system coding / decoding.

Note that low delay coders are often linked to coding speech for duplex telephone links, for example, or to be associated with echo cancellers. Most often designed for sampling frequencies from 8 kHz to 16 kHz, their level of quality is insufficient to code in a manner close to the original of the generic digital audio signals.

The object of the invention is to propose, in this context, a coding system and the associated decoding system which allows, receiver side, to reconstruct both a digital audio signal from quality and a lower quality digital audio signal but whose encoding / decoding delay is as short as possible.

We already know such a coding / decoding system and we will cite the Preprint 4132 of the 99 ^th AES Convention of October 1995 in New York in which Bernhard Grill et al. describe hierarchical digital audio coding systems, that is to say the output bit stream of which comprises a subset of bits which can allow decoding and reconstruction of a significant or relevant sound signal, but of a low quality compared to that obtained by decoding and reconstruction from the total bit stream. Such coding systems include an encoder for coding a high quality sound signal the output of which is connected to the input of a decoder and a difference circuit which makes the difference between the signal obtained at the output of the decoder and the signal d 'origin. The difference signal is itself subjected, in a second stage, to coding, decoding and analogous difference calculation treatments. The third stage codes the residual difference signal. The signals from the encoders of the three stages are then multiplexed so as to form a hierarchical digital stream. Several embodiments are presented, one of which specifies that, in the first stage, the coder is a low bit rate coder which has a relatively low coding delay. The second stage encoder, on the other hand, is a longer delay encoder.

With such a system there are therefore three multiplexed streams in a single output flow, one of these flows generated by the low delay encoder with low delay and lower quality while the other two have higher delays but provide the flow of information necessary for a reproduction of good quality.

However, in the systems presented by Bernhard Grill, each encoder actually consists of a sub-sampled filter bank and an encoder. Likewise, each decoder is actually consisting of a decoder, a bank of filters associated with the bank of encoder and oversampler filters. We have seen that the use of such encoders and decoders in this structure particular entails a still relatively encoding / decoding delay high flow low quality.

The object of the invention is to propose a coding system which has a lower low quality stream coding / decoding delay to the one given by the system described above.

To this end, a coding system according to the invention is characterized in that it comprises a filter bank provided for receive said incoming stream to be encoded and to generate signals respectively in different sub-bands, so-called coders primary coders, for respectively coding said signals into sub-bands and thus form primary streams, decoders receiving said primary streams and decoding said streams, subtractors each of which is intended to make the difference between the signals delivered by the filter bank in a sub-band and the signals from the corresponding decoder, an encoder, called an encoder secondary, to code the signals from the subtractors, and thus generate a secondary flow, and a multiplexer to multiplex the streams into a single global stream primary from primary encoders and the secondary stream from secondary encoder.

It also includes a second bank of filters, known as a secondary filters which receive on each of its inputs the signal of difference from a subtractor and which delivers a filtered flow to the secondary encoder input. Said secondary filter bank advantageously comprises, for each sub-band, an entry for receive the primary stream from the primary encoder and decode by the corresponding decoder to determine, using a model psycho-acoustic, maximum levels of injectable noise in each of the sub-bands, said secondary encoder being an encoder perceptual whose coding is based on psycho-acoustic analysis performed by said secondary filter bank.

According to an alternative embodiment of the invention, said bench secondary filters includes, for each sub-band, an entry for receive the signal in sub-bands from the primary filter bank in order to determine, using a psycho-acoustic model, the maximum levels of injectable noise in each of the sub-bands, said secondary coder being a perceptual coder whose coding is based on the psycho-acoustic analysis carried out by said bench secondary filters.

Advantageously, each primary coder is a coder reconfigurable in debit.

The present invention also relates to a method of multiplexing of a primary frame with a secondary frame generated by a coding system of a signal to be coded, of the type delivering a global flow consisting of a primary flow corresponding to coding of an incoming stream, called primary coding, and of a stream secondary corresponding to secondary coding

It consists in constituting a frame called global frame constituted by the concatenation of a plurality of primary frames and a plurality of fragments of at least one secondary frame, one frame primary alternating with a secondary frame fragment, the number of bits of a secondary frame fragment being equal to the bit rate allocated to the secondary flow multiplied by the duration of transmission of a frame primary. The transmission of global frames is advantageously done all the durations of the primary frames. Similarly, the duration of a frame global is equal to the duration of transmission of a primary frame multiplied by the number of primary frames.

The present invention also relates to a decoding system a stream encoded by a coding system such as the one described above. It includes a flow demultiplexer delivering a plurality of primary streams and a secondary stream, a plurality of primary decoders for decoding said primary streams, the output of each decoder being connected to a corresponding input of a bench primary filters then delivering a low delay decoded stream, the output of each decoder also being connected to an input from one line to corresponding delay whose output is linked to the first input a summator, a secondary decoder delivering a secondary stream decoded supplied to a second input of each adder, the output of each adder being connected to the input of a second filter bank primary to deliver a high quality decoded stream. It comprises in addition a secondary filter bank.

la Fig. 1 est une vue schématique d'un système de codage selon l'invention,

la Fig. 2 illustre le procédé de multiplexage qui est mis en oeuvre dans un système de codage selon l'invention,

la Fig. 3 est un vue schématique d'un système de décodage selon l'invention.

The characteristics of the invention mentioned above, as well as others, will appear more clearly on reading the following description of an exemplary embodiment, said description being made in relation to the accompanying drawings, among which:

Fig. 1 is a schematic view of a coding system according to the invention,

Fig. 2 illustrates the multiplexing method which is implemented in a coding system according to the invention,

Fig. 3 is a schematic view of a decoding system according to the invention.

The coding system shown in FIG. 1 consists of a filter bank 10, the input of which receives an incoming digital audio stream FE to be coded. The filter bank 10 delivers several signals located in different sub-bands, called primary sub-bands. These signals are respectively supplied to the inputs of low-speed primary encoders 20 ₁ to 20 ₄ , here four in number but may be in any number n greater than two. The output of each primary coder 20 _i (i = 1 to n) is connected, on the one hand, to a corresponding input of a multiplexer 30 and, on the other hand, to the input of a low delay primary decoder 40 _i (i = 1 to n). The output of each decoder 40 _i is connected to a first input of a subtractor 50 _i , the other input of which receives the signal from the corresponding primary sub-band delivered by the filter bank 10. The difference signal from the subtractor 50 _i is supplied to the input of a secondary filter bank 60, the output of which is connected to an encoder 70. The output of the encoder 70 is connected to a corresponding input of the multiplexer 30.

Multiplexer 30 interleaves primary flows and secondary respectively from coders 20 and 70. FIG. 2 illustrates the interleaving process.

Two time axes have been shown, one of which is expanded relative to the second, dotted lines showing the time correspondence between these axes. On the first axis, are represented segments whose length corresponds to the duration of establishment t of a primary frame obtained by association of the four primary streams coming from coders 20 ₁ to 20 ₄ . On the other axis, there is shown a global frame TG consisting of a header H, four primary frames TP and four fragments of a secondary frame FTS, the fragments of secondary frame FTS alternating with the primary frames TP. The fragments of secondary frame FTS are the result of a fragmentation of the secondary frame TS delivered by the secondary coder 70. The number of bits of a fragment FTS is equal to the bit rate allocated to the secondary stream multiplied by the duration t of transmission primary coders.

It can be seen that the duration Tt of the global frame TG is an integer multiple of the duration t of the primary frame mentioned above (here four). Likewise, the duration Tt of the global frame TG is an integer multiple of the duration T of the secondary frame TS. Advantageously, the duration of the overall frame Tt is equal to the duration T of a secondary frame TS. In this case, only one secondary frame TS is included in the global frame TG, as is the case in FIG. 2.

It will be noted that the number of primary frames TP and the number of fragments of secondary frames TS ar global frame could be different from four without fundamentally changing the concept of the invention. In particular, this number is not linked to the number of sub-bands contained in a primary frame.

In order to reduce the coding / decoding delay of the primary stream, the emission of the global flow is done all the durations of the frames primary TP. More precisely, for each program, the information of a primary frame TP and of the frame fragment secondary FTS consecutive.

Over the duration Tt of the overall frame, the bit rate allocated to each primary coder 20 _i is variable. This allocation is known to both the coding system and the decoding system. For example, we could decide the allocation according to the energy in each primary sub-band.

The header H contains a synchronization word for setting the decoding system and for delivering the allocations of the different primary coders 20 _i . These frame header allocations sent by the coding system are then used to initialize the decoding system and to remedy any transmission errors.

For each sub-band of the filter bank 10, the filter bank 60 has an input for receiving the concerned sub-band delivered by the primary filter bank 10. From this signal, a model suitable psycho-acoustic, for example the first model proposed by ISO / IEC 13818-3, will determine the maximum noise levels inaudible injectable into each of the sub-bands secondary.

The encoder 70 is a perceptible encoder whose coding is based on the psycho-acoustic analysis provided by the filter bank 60.

When the stream of the primary coder 20 _i has a sufficient number of bits, for example 2.5 bits per sample, it is preferable to replace the original signal at the input of the filter bank for processing according to the psycho-acoustic model by its coded then decoded version delivered by the decoder 40 _i in the primary sub-band considered. The advantage is that the secondary decoder of the decoding system which is associated with the present coding system and which is therefore provided with the same psycho-acoustic model as the filter bank 60 can deduce the fine allocation levels calculated by the secondary coder 70. This saves on transmission costs.

The primary filter bank may be a filter bank of the family of QMF (Quadrature Mirror Filterbank) or benches MOT (Modulated Orthogonal Transforms) filters, with a number enough subbands not to produce a delay too late. A bank of filters modulated in sub-bands of uneven widths or a wavelet type cascaded filter bank or others is also possible, provided that this choice is compatible with the deadline. A filter bank with eight sub-bands modulated from a filter of length thirty two such as the one described by H.S. Malvar in an article entitled "Extended Lapped Transforms: Properties, Applications, and Fast Algorithms "published in IEEE Transactions on signal processing, Vol 40, No 11, pp2703-2714 of November 1992 is a good example of a bench filters adapted to the system of the invention.

Each low delay coder 20 _i can be a coder reconfigurable in bit rate so that the bit rate associated with each sub-band is variable. Each coder 20 _i generates a stream over a small number of grouped samples, representing a constant duration independent of the sub-band. This duration will hereinafter be called the primary duration.

For example, we can choose an LD-CELP (Low Delay - Code Excited Linear Prediction), such as the one described by J.H. Chen and all in an article entitled "A low delay CELP coder for the CCITT 16kb / s speech coding standard "published in IEEE J.Sel.Areas Commun., Vol 10, pp830-849 of June 1992. This LD-CELP coder can contain a choice of dictionaries of different sizes.

With regard to each decoder 40 _i , it will be noted that it could be included in the associated coder 20 _i .

Regarding the secondary filter bank 60, its choice is freer than for the primary filter bank 10 since where there is no constraint on the delay it introduced. Such a filter bank can deliver a variable number of sub-bands by primary sub-band, according to the stationarity of the subband signal. In addition, to overcome overlaps spectrum of the primary filter bank, it is advantageous to use aliasing reduction butterflies, such as those described by 8. Tang and all in an article entitled "Spectral analysis of subband filtered signals "published in ICAASP, Vol 2, pp 1324-1327, 1995.

For example, in the case of a primary filter bank 10 to eight sub-bands, we can choose, for each of the first four primary sub-bands, a bank of MOT type filters (Modulated Orthogonal Transforms) with means allowing, depending on the signal stationarity, switching a window of lengths 128 or 32, producing respectively 64 or 16 sub-bands, and, for the other four primary sub-bands, a type filter bank WORD in 32 sub-bands of length 64.

The bit rate available for the secondary encoder 70 is calculated by subtracting the bit rate used by the low delay primary encoders 20 _i from the total bit rate. For example, for a total bit rate of 64 kbits / s, it would be possible to allocate 32 kbits / s to all of the primary coders 20 ₁ to 20 _n and 32 kbits / s to the secondary coder 70.

The decoding system shown in FIG. 3 is made up of elements whose references are between 110 and 180. Each element is the dual of an element of the coding system shown in FIG. 1 with the exception of elements 180 _i . Its reference is then the same plus a hundred. By way of example, the demultiplexer 130 is the dual of the multiplexer 30.

In the present description, one element is the dual of another element when intended to perform the reverse function of this first.

The decoding system shown in FIG. 3 consists of a demultiplexer 130 whose outputs are respectively connected to the inputs of primary decoders 120 ₁ to 120 ₄ and to a secondary coder 170.

The output of each primary decoder 120 ₁ to 120 ₄ is connected, on the one hand, to an associated delay line 180 ₁ to 180 ₄ and, on the other hand, to an input of a first primary filter bank 110. The output of the filter bank 110 delivers the decoded primary flow Fd. the primary stream decoded Fd is the stream of lower quality but of weak coding / decoding delay.

The output of each delay line 180 ₁ to 180 ₄ is connected to a first input of a corresponding adder 150 ₁ to 150 ₄ .

The output of the secondary decoder 170 is connected to the input of a filter bank 160 whose outputs are respectively connected to the second inputs of the adders 150 ₁ to 150 ₄ .

Finally, the outputs of the adders 150 ₁ to 150 ₄ are respectively connected to the corresponding inputs of a filter bank 110 ′ whose output delivers the high quality decoded stream Fdhq.

A link between each delay line 180 _i and the decoder 170 is provided so as to transmit to the latter, at the desired time, the allocation information present in the primary stream originating from the corresponding decoder 120j.

The demultiplexer 130 of the decoding system realizes the separation of the global frame TG received into primary frames TP and into a secondary frame delivered alternately to the primary decoders 120 ₁ to 120 ₄ and to the secondary decoder 170. The low delay output of the decoding system is obtained by decoding, in the primary decoders 120 _i , primary frames into sub-bands then passing through the reciprocal filter bank 110 of the low-delay filter bank 10. In each of the sub-bands, the primary stream originating from the primary decoder 120 _i and the allocation information it contains are sent to the corresponding 180 _i delay line to supply the high quality part. The allocation information from the delay lines is transmitted, for each primary stream, to the secondary decoder 170 which then performs a decoding of the secondary frame. The reciprocal aliasing reducing butterflies of the coding butterflies are then applied, then the secondary filter bank 160. The signals received from the primary decoders 120 _{i are then added} via the delay lines 180 _i to supply the primary filter bank 110 ' . The high quality Fdhq signal is recovered at the output.

Claims (10)

1. Coding system for a signal which is to be encoded, of the type providing an overall flow consisting of a primary flow corresponding to the encoding of an incoming flow, referred to as the primary flow, and a secondary flow corresponding to the secondary encoding, the period required for the encoding of the said primary encoding being less than that of the secondary encoding, characterised in that it comprises a bank of filters (10) intended to receive the said incoming flow (FE) which is to be encoded, and to generate signals in the different sub-bands respectively, encoders, referred to as primary encoders (20₁ to 20₄), in order, respectively, to encode the said signals in sub-bands and so form the primary flows (TP), decoders (40₁ to 40₄) receiving the said primary flows (TP) and decoding the said flows, subtractors (50₁ to 50₄), each of which is provided in order to implement the differential between the signals delivered by the filter bank (10) in each sub-band and the signals delivered by the corresponding decoder (40₁ to 40₄), an encoder (70), referred to as the secondary encoder, to carry out the encoding ofthe signals deriving from the subtractors (50₁ to 50₄) and accordingly to generate a secondary flow (TS), and a multiplexer (30) to multiplex in one single overall flow (TG) the primary flows (TP) deriving from the primary encoders (20₁ to 20₄) and the secondary flow (TS) deriving from the secondary encoder (70).
2. Coding system according to Claim 1, characterised in that it comprises a second bank of filters (60), referred to as the secondary filter bank, which receives at each of its inputs the differential signal deriving from each subtracter (50₁ to 50₄) and which delivers a filtered flow to the input of the secondary encoder (70).
3. Coding system according to Claim 2, characterised in that the said secondary filter bank (60) comprises, for each sub-band, an intake to receive the primary flow (TP) deriving from the primary encoder (20₁ to 20₄) and decoded by the corresponding decoder (40₁ to 40₄), in order to determine, by means of a psycho-acoustic model, the maximum levels of noise which can be injected into each sub-band, the said secondary encoder (70) being a perceptive encoder, the coding of which is based on the psycho-acoustic analysis effected by the said secondary filter bank (60).
4. Coding system according to Claim 2, characterised in that the said secondary filter bank (60) comprises, for each sub-band, an intake to receive the signal in sub-bands deriving from the primary filter bank (10), in order to determine, by means of a psycho-acoustic model, the maximum levels of noise which can be injected into each of the sub-bands, the said secondary encoder (70) being a perceptive encoder, of which the coding is based on the psycho-acoustic analysis effected by the said secondary filter bank (60).
5. Coding system according to one of the foregoing Claims, characterised in that each primary encoder (20₁ to 20₄) is a encoder which can be reconfigured in its output.
6. Multiplexing process of a primary frame (TP) with a secondary frame (TS), engendered by an encoding system of a signal to be encoded, of the type delivering an overall flow consisting of a primary flow corresponding to an incoming flow, referred to as the primary flow, and a secondary flow corresponding to a secondary encoding, characterised in that it consists of forming a frame, referred to as the overall frame (TG), formed by the concatenation of a plurality of primary frames (TP) and a plurality of fragments (FTS), of at least one secondary frame (TS), a primary frame (TP) alternating with a fragment of the secondary frame (FTS), the number of bits of a fragment of the secondary frame (FTS) being equal to the output effected at the secondary flow (TS) multiplied by the duration of emission of a primary frame (TP).
7. Multiplexing process according to Claim 6, characterised in that the emission of the overall frames (TG) is effected throughout the duration of the primary frames (TP).
8. Multiplexing process according to Claim 6 or 7, characterised in that the duration of an overall frame (TG) is equal to the duration of emission of a primary frame (TP) multiplied by the number of primary frames (TP).
9. System for decoding a flow encoded by a coding system according to one of Claims 1 to 5, characterised in that it comprises a flow multiplexer (130) delivering a plurality of primary flows and a secondary flow, a plurality of primary decoders (120₁ to 120₄) to decode the said primary flows, the output of each decoder (120₁ to 120₄) being connected to a corresponding input of a primary filter bank (110), accordingly delivering a low-speed decoded flow (Fd), the output of each decoder (120₁ to 120₄) being likewise connected to an input of a corresponding delay line (180₁ to 180₄), the output of which is connected to the first input of an adder (150₁ to 150₄), a secondary decoder (170) delivering a secondary decoded flow supplied at a second input of each adder (150₁ to 150₄), the output of each adder (150₁ to 150₄) being connected to the input of a second bank of primary filters (110') to deliver a decoded flow of high quality (Fdqh).
10. Decoding system according to Claim 9, characterised in that it further comprises a secondary bank of filters (160).

Images