Excitation Codebook Research Articles

Dynamic codebook for efficient speech coding based on algebraic codes

A method of encoding a speech signal is presented. This method improves the excitation codebook and search procedure of the conventional Code Excited Linear Prediction (CELP) speech encoders. Use is made of a dynamic codebook (201, 202) based on the combination of two modules: a sparse algebraic code generator (201) associated to a filter (202) having a transfer function varying in time. The generator (102) is a structured codebook with codewords having very few non zero components. The filter (202) shapes the spectral characteristics whereby the resulting excitation codebook (201, 202) exhibits favorable perceptual properties. The search complexity in finding the best codeword is greatly reduced by bringing the search back to the algebraic code domain thereby allowing the sparsity of the algebraic code to speed up the necessary computations.

Speech coding method and system

A speech coding method which can code a speech signal at a bit rate of 8 kb/s or less by a comparatively small amount of calculation to obtain a good sound quality. An autocorrelation of a synthesis signal synthesized from a codevector of an excitation codebook (140) and a linear predictive parameter of an input speech signal is corrected using an autocorrelation of a synthesis signal synthesized from a codevector of an adaptive codebook (120) and a linear predictive parameter and a cross-correlation between the synthesis signal of the codevector of the adaptive codebook (120) and the synthesis signal of the codevector of the excitation codebook (140). A gain codebook (210) is searched using the corrected autocorrelation and a cross-correlation between a signal obtained by subtraction of the synthesis signal of the codevector of the adaptive codebook (120) from the input speech signal and the synthesis signal of the codevector of the excitation codebook (140).

Speech signal coding using correlation values between subframes

A speech signal coding system for coding a speech signal at a bit rate of 8 to 4 kb/s wherein the amount of calculation for fractional search of delays of an adaptive codebook is reduced significantly. Before a fractional delay of the adaptive codebook is found, candidates of integer delay are found by an open-loop using correlation values. A search for a fractional delay by a closed loop is performed for a search range for fractional delays which is provided by ±several samples of each integer delay candidate thus found using the correlation values. The fractional delay search is realized by polyphase filtering of an excitation signal in the past. In the search, a plurality of candidates of fractional delay may be found for each integer delay candidate from the adaptive codebook. In this instance, a fractional delay is determined decisively from the decimal delay candidates after a search of an excitation codebook.

Digital speech coder having optimized signal energy parameters

A speech coder and decoder methodology wherein pitch excitation and codebook excitation source energies (100) are represented by parameters that are readily transmissible with minimal transmission capacity requirements. The parameters are the long term energy value, a short term correction factor which is applied to the long term energy value to match the short term energy, and proportionality factor(s) that specify the relative energy contribution of the excitation sources to the short term energy value (101).

Wear-toll quality 4.8 kbps speech codec

A speech codec operating at low data rates uses an iterative method to jointly optimize pitch and gain parameter sets. A 26-bit spectrum filter coding scheme may be used, involving successive subtractions and quantizations. The codec may preferably use a decomposed multipulse excitation model, wherein the multipulse vectors used as the excitation signal are decomposed into position and amplitude codewords. Multipulse vectors are coded by comparing each vector to a reference multipulse vector and quantizing the resulting difference vector. An expanded multipulse excitation codebook and associated fast search method, optionally with a dynamically-weighted distortion measure, allow selection of the best excitation vector without memory or computational overload. In a dynamic bit allocation technique, the number of bits allocated to the pitch and excitation signals depend on whether the signals are "significant" or "insignificant". Silence/speech detection is based on an average signal energy over an interval and a minimum average energy over a predetermined number of intervals. Adaptive post-filter and the automatic gain control schemes are also provided. Interpolation is used for spectrum filter smoothing, and an algorithm is provided for ensuring stability of the spectrum filter. Specially designed scalar quantizers are provided for the pitch gain and excitation gain.

Glottal source modeling for voice conversion

This paper describes recent advances in glottal source modeling for speech synthesis. In particular two procedures for modeling the glottal excitation waveform are described and applied to voice conversion. One model uses a polynomial to represent the glottal excitation waveform for one pitch period. The coefficients of the polynomial model form a vector that is used to design a glottal excitation code book with 32 entries for voiced excitation. The codebook is designed and trained using two sentences spoken by different speakers. Speech is synthesized using a quantized glottal excitation waveform for one speaker as the excitation for a glottal excitation linear predictive (GELP) synthesizer designed using tract parameters obtained from the speech of another speaker. Our implementation of the LP synthesizer is patterned after both a pitch-excited LP speech synthesizer and a code excited linear predictive (CELP) speech coder. In addition to the glottal excitation codebook, we use a stochastic codebook with 256 entries for unvoiced noise excitation. Analysis techniques are described for constructing both codebooks. The GELP synthesizer, which resynthesizes speech with high quality, provides the speech scientist with a simple speech synthesis procedure that uses established analysis techniques, that is able to reproduce all speech sounds, and yet also has an excitation model waveform that is related to the derivative of the glottal flow and the integral of the residue. Another approach uses the LF glottal volume-velocity waveform to model the characteristics of three voice types: modal, breathy, and vocal fry (creaky). We then convert a modal voice to sound like a breathy or vocal fry voice using the vocal tract characteristics for modal voice and the glottal volume-velocity waveform model for breathy and vocal fry voices as the excitation.

Predictive coding of speech using microphone/speaker adaptation and vector quantization

A problem with speech coders based on trained quantizers is their lack of robustness against variations in the microphone and input filter response. In this paper a simple backward adaptive prefiltering technique is proposed as a means of improving the robustness and quality of a speech coder at no cost in bit rate. The technique is particularly useful in conjunction with vector quantization (VQ) of the linear predictive coding (LPC) parameters. The performance of the prefilter, denoted a microphone and speaker adaptation (MSA) filter, has been evaluated in terms of prediction gain and spectral distortion, together with objective and subjective quality of a 7.5 kbit/s CELP speech coder. In this coder a 10-bit direct VQ of the LPC parameters using the residual energy distortion measure has been applied. This is consistent with the covariance method of LPC analysis. Simulation results illustrate that the MSA filter significantly improves the performance and robustness of the LPC VQ against changes in the input response. The 7.5 kbit/s CELP with a trained excitation codebook and MSA was found to be clearly better (subjectively and objectively) than the one without MSA. The coder with MSA also showed to be practically indistinguishable from the same CELP with unquantized LPC coefficients and a stochastic excitation codebook. >

Method for reducing the search complexity in analysis-by-synthesis coding

A method of encoding speech includes a limited search of a tree-code excitation codebook with a closed loop gain calculation for each test path under consideration. The gain calculation occurs when minimizing an error distance measurement between a synthetic signal defined by each test path being considered and the appropriate speech signal by optimizing a scaling factor of the synthetic signal. The encoding method achieves a significant reduction in computational complexity with minimal loss of performance.

Training method of the excitation codebooks for code-excited linear prediction

This paper considers the codebook for low bit-rate speech coding by code-excited linear prediction (CELP) and the mapping table for the channel code. It also proposes an optimization procedure for sequential training based on the structured excitation codebook. In the discussions of structurization of the codebook, the conjugate structure to represent the output by the sum of two vectors, the pitch synchronized processing, rotating use of the codebook, the time-domain sloped gain, and the integrated mapping to the channel code are considered. The distortion measure used in the training is the same closed-loop measure as is used in the coding. In other words, the distortion between the input speech and the synthesized speech at the perceptually weighted filter output is minimized. By training the codebook, the SNR of the coded speech is improved by approximately 1 dB. The improvement in speech quality is also verified. It is also verified that the robustness against bit error can be improved by the training of the mapping table to the channel code, as well as by training based on the distortion measure considering bit error.

Hybrid switched multi-pulse/stochastic speech coding technique

First Page

Source coding and vector quantization with codebook-excited neural networks

The quantization property of layered neural networks is studied in this paper. We first review the layered neural network-based coders or quantizers developed in recent years and show that their poor performances is due to their independent training schemes for each component in the coder. Then an alternative model named a codebook-excited neural network is proposed, where an encoded vector is approximated by the output of the network driven by one vector selected from an excitation codebook. The network and the excitation codebook are jointly trained with the error back-propagation algorithm. Simulations with a Gauss-Markov source demonstrate that the quantization performance of the codebook-excited feedforward neural network is not worse than that of the connectionist vector quantizer formed by a set of single-layer neural units which satisfied the optimal quantization conditions, and that the performance of the codebook-excited recurrent neural network is very close to the asymptotic performance bound of block quantizers. The codebook-excited neural network is applicable with any distortion measure. For a zero-mean, unit variance, memory-less Gaussian source and a squared-error measure, a 1 bit/sample two-dimensional quantizer with a codebook-excited feedforward neural network is found always to escape from the local minima and converge to the best one of the three local minima which are known to exist in the vector quantizer designed using the LBG algorithm. Moreover, due to its conformal mapping characteristic, the codebook-excited neural network can be applied to designing the vector quantizer with any required structural form on its codevectors.

Finite state CELP for variable rate speech coding

The performance of a variable rate code excited linear predictor system is investigated. The coding system is based on a finite state CELP (FSCELP) frame work. Each individual state is primarily identified with a LPC model order, LPC coefficients bit allocation, excitation code book population density and state encoding rate. Successive input speech vectors are encoded at a rate that depends on the current state of the FSCELP system and the input vector characteristics. The use of a finite state system involves implicit clustering of speech signals. The lower rate states are selected during highly correlated steady state speech segments when relatively few bits are required to obtain adequate fidelity. For speech signals with a strong glottal excitation, unvoiced signals and transient speech segments, a relatively greater quantisation accuracy is needed to obtain good fidelity and therefore higher rate states of the system are used. Further improvement is obtained by using gamma populated excitation codebooks, for those states that are mainly used to encode speech signals with a strong underlying glottal excitation pulses. Experiments focus on investigation of the varying encoding requirements of the excitation signal for low pass, voiced, unvoiced and transient speech signals. The parameters of the finite state CELP system are designed to match the encoding requirements of typical speech signals. The greater part of the coding gain is obtained from variable rate encoding of the excitation signal. Using a six-state FSCELP, good quality speech is obtained at an average, maximum and minimum bit rates of 4 kbit/s, 10 kbit/s and 2 kbit/s, respectively.

Improving performance of code excited LPC-coders by joint optimization

This article describes a CELP speech coding algorithm where the coder parameters are jointly optimized. For that, the relation between pitch period, pitch predictor coefficient, codebook entry and scaling factor is derived. Different approaches to solve this optimization problem with low computational load are discussed. It is shown that the length of the excitation codebook can be reduced significantly compared to the sequential optimization of the coder parameters. The influence of the parameter quantization is examined and an estimation of the computational load is given.

Binary code excited linear prediction (BCELP): new approach to celp coding of speech without codebooks

A new analysis-by-synthesis speech coding approach able to produce good quality speech in the vicinity of 4.8 kbit/s is presented. The new approach produces the same speech quality as obtained by CELP codecs without needing any excitation codebook storage. The new coder employs a very simple excitation search procedure and possesses an inherent robustness against channel errors. The approach is based on the ternary code excitation CELP introduced in Reference 3.

Classified vector excitation speech coding

Code excited linear prediction (CELP) has been shown [1] to be capable of achieving high quality speech compression at very low bit rates. However, the original method suffers from the two drawbacks that it requires a large amount of computation, and the Gaussian codebook used is not able to reproduce a sharp impulse in the coded residual waveform. The authors present a classified vector excitation coding (CVXC) scheme based on selective modelling of the LPC residual which attempts to overcome both these drawbacks by using separate codebooks for separate excitation types. A further feature of our coder which helps preserve the nature of pulsive residuals, is the use of pairwise nearest neighbour (PNN) matching [2] to generate the excitation codebooks.

On improving voice periodicity prediction in codebook-excited LPC coders

Codebook-excited predictive coders are able to synthesize high-quality speech at bit rates of 8 kbits/s and above. However, at lower bit rates (e.g. 4.8 kbits/s), speech enhancement techniques, such as adaptive post-filtering, are needed to improve subjective performance [cf. P. Kroon and B. S. Atal, Proc. ICASSP (1987)]. The main source of quality degradation at 4800 b/s is due to the reduced transmission bandwidth of the excitation parameters, which results in an inaccurate representation of the LPC innovation signal. In this paper, methods are proposed for improving the voiced speech excitation and prediction in the codebook-excited coder. The modification uses a composite (i.e., product) excitation codebook together with a close-loop pitch prediction filter that is optimized relative to the source codebook. Several optimization creiteria for the source and predictor are compared in terms of performance and computational complexity. The best result shows a segmental SNR improvement of nearly 2 dB over the conventional stochastically excited LPC coder.