J.-P. Adoul

Fast CELP coding based on algebraic codes

Code-Excited Linear Prediction (CELP) produces high quality synthetic speech at low bit rate. However the basic scheme leads to huge computational loads. The paper describes a related scheme, which allows real time implementation on current DSP chips. The very efficient search procedure in the codebook is achieved by means of a new technique called backward filtering and the use of algebraic codes. RSB performances are reported for a variety of conditions.

16 kbps wideband speech coding technique based on algebraic CELP

The application of algebraic code excited linear prediction (ACELP) coding to wideband speech is presented. An algebraic codebook with a 20 bit address can be used without any storage requirements and, more importantly, with a very efficient search procedure which allows for real-time implementation. The authors describe an efficient procedure for searching such a large codebook deploying a focused search strategy, where less than 0.1% of the codebook is searched with performance very close to that of a full search. High-quality speech at a bit rate of 13 kbps was obtained.<<ETX>>

A toll quality 8 kb/s speech codec for the personal communications system (PCS)

A toll quality speech codec at 8 kb/s suitable for the future personal communications system is presented. The codec is currently under standardization by the ITU-T (successor of CCITT) where the codec terms of reference were mainly determined considering PCS application. The encoding algorithm is based on algebraic code-excited linear prediction (ACELP) and has a speech frame of 10 ms. Efficient pitch and codebook search strategies, along with efficient quantization procedures, have been developed to achieve toll quality encoded speech with a complexity implementable on current fixed-point DSP chips. Formal subjective listening tests, performed by ITU-T SG 12, showed that the codec quality is equivalent to that of G.726 ADPCM at 32 kb/s in error-free conditions and it outperforms G.726 under error conditions. The codec performs adequately under tandeming conditions, and can support a frame erasure rate up to 3% with a degradation in its performance that is still worse than the ITU-T requirements, and this is one subject of study for the next phase. The algorithm has been implemented on a single fixed-point DSP for the ITU-T subjective rest, and required about 29 MIPS. An optimized version, however, requires 24 MIPS without any speech quality degradation. >

On reducing computational complexity of codebook search in CELP coder through the use of algebraic codes

A general framework is introduced which allows both fast search and freedom in designing codebooks with good statistical properties. Several previously proposed schemes are compared from this viewpoint. A backward filtering formulation is given to show that sparse algebraic codes (SACs) (i.e., with few nonzero components) offer distinct advantages. It is shown that they reduce the optimal-search computation per codeword. They also allow control of the statistical properties of the codebook in the time and frequency domains. This control can be dynamic in the sense that it can be made to evolve as a function of the linear predictive coding model A(z). The algebraic-code excited linear prediction (ACELP) technology which allows full duplex operation on a single TMS320C25 at rates between 4.8 and 16 kb/s and which is based on SAC-driven dynamic codebooks is described.<<ETX>>

A comparison of some algebraic structures for CELP coding of speech

The paper discusses how binary error-correcting codes can be used to provide (+1/-1)-waveform codebooks that speed up search in CELP vocoders. Four coding techniques operating at half bit per sample with respectively 8, 16, 24 and 32 samples are compared in terms of complexity and SNR performance, Recent results on spherical codes from regular point lattices are also reported.

Baseband speech coding at 2400 bps using "Spherical vector quantization"

This paper concerns a new Quantization Scheme for efficient encoding of waveforms below or about one bit per sample. This technique is then applied to the encoding of baseband residual signals to demonstrate the feasibility of (baseband) Residual Excited LPC at 2400 hit/sec. The technique called Spherical Vector Quantization is described in which a block of n consecutive samples is quantized as a vector. The magnitude of the vector is transmitted independently of the vectors orientation. This vectors orientation is vector quantized using a codebook which can be seen as representing a set of N points on a unit hypersphere. The cases for blocks of n = 8 and 24 are discussed which make use of results by Conway and Sloane on regular point lattices. For n = 8, the algorithm is detailed which solves both the problem of finding the closest point on the hypersphere and the problem of determining the index of that point. Experiments involving SVQ in coding the speech baseband residual are described which show in particular that subband coding does not contribute any quality improvement when SVQ is used.

Fast CELP coding based on the Barnes-Wall lattice in 16 dimensions

The authors present an algebraic code-excited linear prediction (CELP) speech coder where the innovation codebook comes from the first spherical code of the Barnes-Wall lattice in 16 dimensions. Novel fast optimal algorithms for finding the best sequence in this Barnes-Wall shell innovation codebook are described. This algebraic codebook makes it possible to design a CELP coder at 9.6 kb/s with good quality and still implementable on a current digital-signal-processing chip.<<ETX>>

8 kbit/s coding of speech with 6 ms frame-length

The coder utilizes backward adaptation for updating the LP (linear prediction) parameters, and improved forward pitch analysis. A novel frequency domain approach called transform coded excitation is used for efficient quantization and encoding of the innovative excitation signal. The coder uses a combination of time-domain (linear prediction and pitch analysis) and frequency-domain (transform coding) techniques to achieve the best reproduction of the original signal in the perceptual sense. Good speech quality was obtained with this approach.<<ETX>>

Low complexity LSF quantization for wideband speech coding

State-of-the-art narrowband speech coders operating from 4 to 16 kbit/s are mostly based on the code-excited linear predictive (CELP) model. They achieve a good synthesis quality usually at the expense of a high coding complexity. For example, in the 8 kbit/s G.729 coder the innovation codebook search is responsible for approximately half the total coder complexity, the latter being close to 20 MIPS in fixed-point DSP implementation. Less known is the relative part of spectral quantization, which is around 8% of the total complexity. CELP coders are still relevant for wideband speech coding but their complexity is greater than in the narrowband case, which becomes critical for real-time implementations. We propose in this article a two-stage algebraic-stochastic line spectral frequency (LSF) quantization scheme. It combines the strengths of algebraic and stochastic techniques, namely low computation and storage cost and good performance. The generalized Lloyd-Max algorithm is adapted for optimizing lattice codebooks obtained by spherical truncation. Simulations with a Gaussian source show that the quantization method exhibits good quality/complexity tradeoffs. Several stochastic-algebraic LSF quantizers are derived and compared to a more conventional technique.

Real-time implementation of a 9.6 kbit/s ACELP wideband speech coder

The real-time implementation of a wideband ACELP speech coder at 9.6 kb/s is presented. The coder is implemented on a TMS320C30 floating-point DSP chip. The attempt to implement an ACELP coder for wideband speech in real time results in 3-4 times more complexity than that for narrowband speech. Very efficient algorithms for searching the pitch and codebook parameters have been introduced. The pitch search was brought down to 20% of real time by the combination of an efficient open-loop approach and a decimation procedure. The excitation search complexity was significantly reduced by using two codebooks. The first models the main features in the excitation and is very efficiently searched using focused search. The second has a simple structure and does not need exhaustive search. The quality of the encoded wideband speech at 9.6 kb/s was judged vastly superior to that of the original narrowband speech.<<ETX>>