Coded data generation or conversion – Digital code to digital code converters
Reexamination Certificate
2000-03-17
2002-04-09
Phan, Trong (Department: 2818)
Coded data generation or conversion
Digital code to digital code converters
C341S067000
Reexamination Certificate
active
06369722
ABSTRACT:
BACKGROUND OF THE INVENTION
This invention relates to the coding of digital signals and more especially coding, decoding and where applicable transcoding methods using vector quantification.
Vector quantification is commonly used in various coding methods. In what follows in this description, attention will be given more particularly to the coding of audio signals, though this is non-limiting.
The invention considers the problem of scalability of the digital data stream transmitted between a coder and a decoder. This property consists in the ability of the coder to construct variable rate output data streams on the basis of the same coding scheme applied to the coded digital signal, and in the corresponding ability of the decoder to reconstruct a faithful version of the signal.
The difficulty here is to secure the highest possible coding quality for each data rate value without unduly increasing the complexity of the circuits used.
Data stream scalability is of particular importance where the data stream is likely to be carried on packet switching networks, such as networks operating according to the IP (Internet Protocol). Historically, the majority of coders have been developed for broadcasting or communications applications in circuit mode, leading to fixed rate coders or coders with the rate selected from several possible values when the connection is set up. In the packet mode context, it is better that the rate should vary more dynamically, so that the data stream can be matched to the congestion encountered when the packets are conveyed while ensuring that the communication is maintained.
Patent application WO99/03210 describes a scalable data stream audio coder having several coding stages. The first stage comprises a coding core, which may be a standard coder of the CELP (Code-Excited Linear Prediction) or other type, and possibly a bank of sub-band separation filters and additional modules to quantify sub-bands higher than the one processed by the coding core. The second stage comprises another bank of filters to separate into sub-bands the residual error signal at the output of the coding core, modules to quantify the components coming from this separation into sub-bands and where applicable additional modules to quantify the residual error signals at the output of the modules quantifying the higher sub-bands of the first stage. The residual error signals at the output of the second stage can then be processed by a third stage etc. The output digital data stream from this coder can be scaled by adjusting the number of coding/quantification modules contributing to it. For the lowest rate, only the coding core is involved. To increase the rate, components and/or stages quantifying the residual errors are added,
U.S. Pat. No. 5,864,650 discloses a method for generating a library of the tree-structured type containing 2
L
vectors obtained by algebraic sums of L basic vectors. As regards the coder, the search for the vector coding a specific value of a signal to be transmitted proceeds by successive selections from the root to the leaves of the code tree, stopping at the code tree stage corresponding to the number of quantification bits available. A library of this kind is however very constrained and does not allow good coding quality.
U.S. Pat. No. 4,899,384 discloses a bit allocation method in connection with coding in sub-bands, this allocation taking account of the level and energy distribution in the band and in each sub-band, so as to comply with a global data rate constraint while keeping degradation of the signal transmitted to a minimum. However, the quantification envisaged in this document is scalar quantification and not vector quantification.
One main purpose of this invention is to obtain fine adjustment of the rate in the scalability of the digital data stream, allowing the best rate/quality compromise to be sought as a function of the communication conditions.
SUMMARY OF THE INVENTION
A first aspect of the invention refers to a method for coding a signal in which, on the basis of a portion of the signal, at least one coding parameter vector is selected belonging to a reference library containing 2
Q
vectors each designated by an address of Q bits, and an output digital data stream is formed containing an index deduced from the address of the vector selected from the reference library. According to this method, the output digital data stream has a rate which is adjustable by varying the number Q−p of bits forming this index contained in the digital data stream. For each rate value corresponding to a value of the integer p≧0, the index contained in the digital data stream is formed of Q−p bits which, completed by p arbitrary bits of predetermined positions, define the indices of Q bits representing a group of 2
p
addresses including the address of the vector selected from the reference library.
A reduction in the coder rate results from the deletion of p bits of certain vector quantification indices. By varying this number p, we are able to achieve very fine scalability of the output digital data stream.
In order to minimise the impact of the reduction in rate on the quality of the signal which the decoder is capable of reconstructing, the p bits deleted will generally be the least significant bits of the index concerned.
Provision can then be made, for least one rate value corresponding to a value of the integer p>0, for the reference library to be ordered so as to verify a criterion of minimum distances between the vectors belonging to each group of 2
p
vectors of the reference library having respective addresses which differ only in the p bits having the predetermined positions mentioned.
Another possibility is to look up a transcoding table on the basis of the address of the vector selected from the reference library to obtain a first index of Q bits. For each rate value corresponding to a value of the integer p≧0, the index contained in the output digital data stream is then formed of Q−p bits extracted from this first index of Q bits excluding p bits of predetermined positions.
This last possibility is of interest in that it enables an existing coder, optimised according to other criteria, to be matched to the concept of scalability, the transcoding table enabling the degradation in quality due to the removal of certain bits to be limited. Under these conditions, the coder may have a first mode of operation in which, for each rate value corresponding to a value of the integer p≧0, the index contained in the output digital data stream is formed of Q−p bits extracted from this first index of Q bits and a second mode of operation with defined rate in which the address of the vector selected from the reference library is included in the output digital data stream instead of this index formed of Q−p bits. One or other of the two modes will be chosen depending on the capabilities of the decoder receiving the data stream and/or according to the means used to convey the data stream as far as this decoder.
A second aspect of the invention relates to a method for decoding a digital data stream representative of a coded signal, the digital data stream containing, for a portion of the signal, at least one index enabling a coding parameter vector, used to construct a decoded version of this portion of the signal, to be obtained by means of a reference library containing 2
Q
vectors each designated by an address of Q bits. In this decoding method, the digital data stream rate depends on the number Q−p of bits forming this index. For each rate value corresponding to a value of the integer p≧0, we obtain the coding parameter vector on the basis of at least one vector in the reference library having an address which belongs to a group of 2
p
addresses represented respectively by the indices of Q bits formed by completing by p arbitrary bits of predetermined positions the index of Q−p bits contained in the digital data stream.
Several methods may be applied to obtain the coding parameter vector on the bas
Lockwood Philipe
Murgia Carlo
Richard Gaël
Matra Nortel Communications
Phan Trong
Trop Pruner & Hu P.C.
LandOfFree
Coding, decoding and transcoding methods does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Coding, decoding and transcoding methods, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Coding, decoding and transcoding methods will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2825143