Abstract: Methods of designing successively refinable Trellis-Based Scalar-Vector quantizers (TB-SVQ) include a multi-stage process wherein a TB-SVQ is applied to a set of digital data to set up a codebook boundary and to obtain a non-uniform density gain for a constellation in which the data signals will be encoded. In at least one more stage, a Trellis coded quantizer (TCQ) is applied to the output codebook boundary of the first stage to obtain a granular or shaping gain of 1.53 dB. The inventive methods successively refine the TB-SVQ so that robust signal transmission is achieved. By applying a multi-stage process wherein a TB-SVQ is utilized in the first stage and a TCQ is utilized in the second and successive stages, the computational complexity and time for encoding the constellation are greatly reduced.

Abstract: A method (and apparatus) for coding an audio signal, the method comprising the steps of partitioning the audio signal into a sequence of successive frames; calculating one or more noise thresholds for each of a plurality of frames in the sequence, each noise threshold for a particular one of the frames corresponding to a different perceptual coding quality for the particular frame; estimating a bit demand for each of a corresponding one or more perceptual coding qualities for each frame, wherein each estimated bit demand comprises a number of bits which would be used to code a given frame at the corresponding perceptual coding quality; selecting one of the perceptual coding qualities for the coding of a particular frame based upon the estimated bit demand for the perceptual coding quality for the particular frame, and further based on one or more bit demands estimated for one or more other frames; and coding the particular frame based on the noise threshold corresponding to the selected perceptual coding qual

Abstract: Encoding of prototype waveform components applicable to telecommunication systems provides improved voice quality enabling a dual-channel mode of operation which permits more users to communicate over the same physical channel. A prototype word (PW) gain is vector quantized using a vector quantizer (VQ) that explicitly populates a codebook by representative steady state and transient vectors of PW gain for tracking the abrupt variations in speech levels during onsets and other non-stationary events, while maintaining the accuracy of the speech level during stationary conditions.

Abstract: With respect to each of codes corresponding to code vectors in a code book stored in a code book storage section 82, an expectation degree storage section 84 stores an expectation degree at which observation is expected when an integrated parameter with respect to a word as a recognition target is inputted. A vector quantization section 81 vector-quantizes the integrated parameter and outputs a series of codes of a code vector which has a shortest distance to the integrated parameter. Further, a chi-square test section 83 makes a chi-square test with use of the series of codes outputted from the vector quantization section 81 and an expectation degree of each code stored in the expectation degree storage section 84, thereby to obtain properness as to whether or not the integrated parameter corresponds to a recognition target. Further, recognition is performed, based on the chi-square test result. As a result of this, recognition can be performed without considering time components which a signal has.

Abstract: A variable bit-rate speech coding method determines for each subframe a quantised vector d(i) comprising a variable number of pulses. An excitation vector c(i) for exciting LTP and LPC synthesis filters is derived by filtering the quantised vector d(i), and a gain value gc is determined for scaling the pulse amplitude excitation vector c(i) such that the scaled excitation vector represents the weighted residual signal {tilde over (s)} remaining in the subframe speech signal after removal of redundant information by LPC and LTP analysis. A predicted gain value ĝc is determined from previously processed subframes, and as a function of the energy Ec contained in the excitation vector c(i) when the amplitude of that vector is scaled in dependence upon the number of pulses m in the quantised vector d(i). A quantised gain correction factor {circumflex over (&ggr;)}gc is then determined using the gain value gc and the predicted gain value ĝc.

Abstract: A code book is a set of code vectors to be selected when linear predictive parameters are vector-quantized as for an input audio which is divided into frames, each of which is further divided into sub frames.

Abstract: A method and apparatus for coding a quasi-periodic speech signal. The speech signal is represented by a residual signal generated by filtering the speech signal with a Linear Predictive Coding (LPC) analysis filter. The residual signal is encoded by extracting a prototype period from a current frame of the residual signal. A first set of parameters is calculated which describes how to modify a previous prototype period to approximate the current prototype period. One or more codevectors are selected which, when summed, approximate the error between the current prototype period and the modified previous prototype. A multi-stage codebook is used to encode this error signal. A second set of parameters describe these selected codevectors. The decoder synthesizes an output speech signal by reconstructing a current prototype period based on the first and second set of parameters, and the previous reconstructed prototype period.

Abstract: A complete system and method for accurate and robust speech recognition based on the application of three perceptual processing techniques to the speech Fourier spectrum to achieve a robust perceptual spectrum and the accurate recognition of that perceptual spectrum by projecting the perceptual spectrum onto a set of reference vowel spectrum vectors for input to a speech recognizer. The invention comprises a perceptual speech processor for preceptually processing the input speech spectrum vector to generate a perceptual spectrum, a storage device for storing a plurality of reference spectrum vectors and a phonetic feature mapper, coupled to said perceptual speech processor and to said storage device, for mapping said perceptual spectrum onto said plurality of reference spectrum vectors.

Abstract: With respect to each of codes corresponding to code vectors in a code book stored in a code book storage section, an expectation degree storage section stores an expectation degree at which observation is expected when an integrated parameter with respect to a word as a recognition target is inputted. A vector quantization section vector-quantizes the integrated parameter and outputs a series of codes of a code vector which has a shortest distance to the integrated parameter. Further, a chi-square test section makes a chi-square test with use of the series of codes outputted from the vector quantization section and an expectation degree of each code stored in the expectation degree storage section, thereby to obtain properness as to whether or not the integrated parameter corresponds to a recognition target. Further, recognition is performed, based on the chi-square test result. As a result of this, recognition can be performed without considering time components which a signal has.

Abstract: A device capable of achieving recognition at a high accuracy and with fewer calculations and which utilizes an HMM. The present device has a vector quantizing circuit generating a model by quantizing vectors of a training pattern having a vector series, and converting the vectors into a label series of clusters to which they belong, a continuous distribution probability density HMM generating circuit for generating a continuous distribution probability density HMM from a quantized vector series corresponding to each label of the label series, and a label incidence calculating circuit for calculating the incidence of the labels in each state from the training vectors classified in the same clusters and the continuous distribution probability density HMM.

Abstract: An apparatus for automatically rendering both encoded and unencoded data files comprises a data processor and a decoder. The data processor receives a set of input data from a storage medium, such as an optical storage medium, and determines whether the set of input data is encoded or unencoded. If the input data is encoded, then the data processor provides to the decoder a set of encoded data, and generates an indication signal to cause the decoder to be activated. In response, the decoder decodes the set of encoded data and provides as output a set of decoded data. This decoded data may then be provided to a signal transport mechanism, such as a digital signal bus, to be rendered by a digital device such as a computer, or to a digital to analog converter which converts the decoded data into analog signals for driving an analog device, such as a speaker.

Abstract: A random code vector reading section and a random codebook of a conventional CELP type speech coder/decoder are respectively replaced with an oscillator for outputting different vector streams in accordance with values of input seeds, and a seed storage section for storing a plurality of seeds. This makes it unnecessary to store fixed vectors as they are in a fixed codebook (ROM), thereby considerably reducing the memory capacity.

Abstract: Compression of speech may be achieved through the adaptive generation of a compressed sound output. A first processing element may be used to characterize a first sound representation such that a first characterization result is produced. A comparison element may be provided to compare a first comparison input that is related to the first sound representation with a second comparison input that is related to the first characterization result. A determination may be made on whether further processing is desirable based on whether the first comparison result satisfies a first predetermined threshold criteria. Additionally, a second processing element may be included to characterize a second sound representation and to produce a second characterization result only if the first comparison result satisfies the first predetermined threshold. A compressed sound output is generated whose contents are determined based on at least the first comparison result.

Abstract: Encoding of prototype waveform components applicable to GeoMobile and Telephony Earth Station (TES) providing improved voice quality enabling a dual-channel mode of operation which permits more users to communicate over the same physical channel. A prototype word (PW) gain is vector quantized using a vector quantizer (VQ) that explicitly populates the codebook by representative steady state and transient vectors of PW gain for tracking the abrupt variations in speech levels during onsets and other non-stationary events, while maintaining the accuracy of the speech level during stationary conditions. The rapidly evolving waveform (REW) and slowly evolving waveform (SEW) component vectors are converted to magnitude-phase. The variable dimension SEW magnitude vector is quantized using a hierarchical approach, i.e., a fixed dimension SEW mean vector computed by a sub-band averaging of SEW magnitude spectrum, and only the REW magnitude is explicitly encoded.

Abstract: A system controller (106) includes a speech encoder (107) that dynamically segments frames of a low bit rate digital voice message. Speech model parameters have been generated in a sequence of frames. The speech model parameters include quantized speech spectral parameter vectors. The speech encoder selects (1820) a first quantized speech spectral parameter vector as a current anchor vector, selects (1820, 1830) a second quantized speech spectral parameter vector located a predetermined number of frames (LMAX) from the current anchor vector as a target speech parameter vector, and perturbs (1840) the target speech parameter vector to derive a plurality (K) of perturbed speech parameter vectors.

Abstract: An excitation vector generator comprises a pulse vector generating section having N channels (N≧1) for generating pulse vectors, a storing section for storing M (M≧1)kinds of dispersion patterns every channel in accordance with N channels, a selecting section for selectively taking out a dispersion pattern from the storing section every channel, a dispersion section for performing a superimposing calculation of the extracted dispersion pattern and the generated pulse vectors every channel so as to generate N dispersion vectors, excitation vector generating section for generating an excitation vector from N dispersion vectors generated.

Abstract: A method and an access mechanism for determining the storage address of a predetermined data value (D1, D2, D3) in a memory device is disclosed. The data values are stored in an increasing order sequentially in a column direction according to a binary tree data structure. A new subtree root node (B(L(X)), B(R(X)), A=1 is calculated from the previous leaf node address (LN) when the data value to be searched is not located in the previous subtree. Since a new subtree root node (X1) is always calculated from a previous leaf node address and the comparison result between the searched and read out value, the number of row address changes can be kept to a minimum whilst a high speed for the subtree searching is maintained. The search method and the access means is memory efficient since no pointers are used and fast, since the address of a next memory location to be investigated can always be calculated from the previous address and the last comparison result.

Abstract: A multi-channel signal encoder includes an analysis part with an analysis filter block having a matrix-valued transfer function with at least one non-zero non-diagonal element. The corresponding synthesis part includes a synthesis filter block (12M) having the inverse matrix-valued transfer function. This arrangement reduces both intra-channel redundancy and inter-channel redundancy in linear predictive analysis-by-synthesis signal encoding.

Abstract: A speech coder for high quality coding speech signals at low bit rates is disclosed. An excitation quantization unit 12 expresses an excitation signal in terms of a combination of a plurality of pulses. A codebook (i.e., an amplitude codebook 13) collectively quantizes either amplitude or position of pulses, and executes excitation signal quantization other parameter by making retrieval of the codebook.

Abstract: A method and apparatus for reducing the complexity of linear prediction analysis-by-synthesis (LPAS) speech coders. The method and apparatus include product code vector quantization (PCVQ) of multi-tap pitch predictor coefficients, which reduces the search and quantization complexity of an adaptive codebook. The pitch predictor vector quantizes the predictor parameters using at least two codebooks, which are effectively subcodebooks of the pitch predictor adaptive codebook. Further included is a procedure for generating and selecting code vectors consisting of ternary (1,0,−1) values, for optimizing a fixed codebook. The fixed codebook makes a single pass derivation of pulse position in the excitation signal. Serial optimization of the adaptive codebook first and then the fixed codebook, produces a low complexity LPAS speech coder of the present invention.

Abstract: A method and apparatus for interleaving line spectral information quantization methods in a speech coder includes quantizing line spectral information with two vector quantization techniques, the first technique being a non-moving-average prediction-based technique, and the second technique being a moving-average prediction-based technique. A line spectral information vector is vector quantized with the first technique. Equivalent moving average codevectors for the first technique are computed. A memory of a moving average codebook of codevectors is updated with the equivalent moving average codevectors for a predefined number of frames that were previously processed by the speech coder. A target quantization vector for the second technique is calculated based on the updated moving average codebook memory. The target quantization vector is vector quantized with the second technique to generate a quantized target codevector.

Abstract: A speech signal is encoded using code excited linear prediction for use in transmitting the speech signal to a receiver. The speech signal is sampled. A current sample of the speech signal is predicted based on in part a previous sample. An innovation sequence is determined based on in part a prediction error between the predicted current sample and the current sample of the speech signal. A code from each of a plurality of codebooks is selected. A combination of the selected codes is the determined innovation sequence. An index of the selected codes is identified and transmitted to the receiver. The transmitted index enables reconstruction of the speech signal at the receiver.

Abstract: Quantization unit (108) comprises evaluator (120) and comparator (122) in signal processing for identifying an utterance in system (100). The evaluator (120) weights a first intermediate result of an operation on a first set of a plurality of speech parameters (104) differently than a second intermediate result of an operation on a second set of the plurality of speech parameters (104) in a weighted representation of the plurality of speech parameters (104). The comparator (122) employs the weighted representation of the plurality of speech parameters (104) to determine a vector index to represent the plurality of speech parameters (104). The quantization unit (108), in one example, can employ split vector quantization in conjunction with the weighted representation to determine a vector index to represent the plurality of speech parameters (104).

Abstract: A constraint relieving section receives an input vector and transforms it with a predetermined transform function to generate a target vector obtained by subtracting a constraint vector representing a predetermined constraint from the input vector. An error evaluation section selects from a codebook a code vector constituting a quantization vector having a minimum error with respect to the target vector and outputs an index representing this code vector.

Abstract: A method and system for reduction of quantization-induced block-discontinuities arising from lossy compression and decompression of continuous signals, especially audio signals. One embodiment encompasses a general purpose, ultra-low latency, efficient audio codec algorithm. More particularly, the invention includes a method and apparatus for compression and decompression of audio signals using a novel boundary analysis and synthesis framework to substantially reduce quantization-induced frame or block-discontinuity; a novel adaptive cosine packet transform (ACPT) as the transform of choice to effectively capture the input audio characteristics; a signal-residue classifier to separate the strong signal clusters from the noise and weak signal components (collectively called residue); an adaptive sparse vector quantization (ASVQ) algorithm for signal components; a stochastic noise model for the residue; and an associated rate control algorithm.

Abstract: A set of speaker dependent models or adapted models is trained upon a comparatively large number of training speakers, one model per speaker, and model parameters are extracted in a predefined order to construct a set of supervectors, one per speaker. Dimensionality reduction is then performed on the set of supervectors to generate a set of eigenvectors that define an eigenvoice space. If desired, the number of vectors may be reduced to achieve data compression. Thereafter, a new speaker provides adaptation data from which a supervector is constructed by constraining this supervector to be in the eigenvoice space based on a maximum likelihood estimation. The resulting coefficients in the eigenspace of this new speaker may then be used to construct a new set of model parameters from which an adapted model is constructed for that speaker. The adapted model may then be further adapted via MAP, MLLR, MLED or the like.

Abstract: In an encoding system, input audio data are subjected to vector quantization in accordance with a content of a code book to produce indexes. In addition, the encoding system produces a number of new pattern vectors which are suited to the audio data. Based on a result of the vector quantization, the encoding system determines a replacement candidate for a pattern vector which is registered with the code book. With respect to each of the new pattern vectors, code book renewal data are produced to represent a linear combination of pattern vectors which are registered with the code book other than the replacement candidate. Herein, the linear combination of pattern vectors has a distance which is the closest to the new pattern vector. Then, the replacement candidate of the pattern vector of the code book is replaced with a new pattern vector for replacement which is newly created based on the code book renewal data. Thus, it is possible to renew the content of the code book.

Abstract: A channel optimized vector quantization apparatus includes a device for weighting a sample vector x by a weighting matrix A and a device for weighting a set of code book vectors ĉr by a weighting matrix B. Device form a set of distance measures {dw(Ax,Bĉr)} representing the distance between the weighted sample vector Ax and each weighted code book vector Bĉr. Other device form a set of distortion measures {&agr;i(x)} by multiplying each distance measure by a channel transition probability Pr|i that an index r has been received at a decoder when an index i has been sent from an encoder and adding together these multiplied distance measures for each possible index r. Finally device determine an index imin corresponding to the smallest distortion measure &agr;i(x) and represents the sample vector by this index imin.

Abstract: A speech coder and a method for speech coding wherein the speech signal is represented by an excitation signal applied to a synthesis filter. The speech is partitioned into frames and subframes. A classifier identifies which of several categories the speech frame belongs to, and a different coding method is applied to represent the excitation for each category. For some categories, one or more windows are identified for the frame where all or most of the excitation signal samples are assigned by a coding scheme. Performance is enhanced by coding the important segments of the excitation more accurately. The window locations are determined from a linear prediction residual by identifying peaks of the smoothed residual energy contour. The method adjusts the frame and subframe boundaries so that each window is located entirely within a modified subframe or frame.

Abstract: Tonal audio signals can be modeled as a sum of sinusoids with time-varying frequencies, amplitudes, and phases. An efficient encoder and synthesizer of tonal audio signals is disclosed. The encoder determines time-varying frequencies, amplitudes, and, optionally, phases for a restricted number of dominant sinusoid components of the tonal audio signal to form a dominant sinusoid parameter sequence. These components are removed from the tonal audio signal to form a residual tonal signal. The residual tonal signal is encoded using a residual tonal signal encoder (RTSE). In one embodiment, the RTSE generates a vector quantization codebook (VQC) and residual codebook sequence (RCS). The VQC may contain time-domain residual waveforms selected from the residual tonal signal, synthetic time-domain residual waveforms with magnitude spectra related to the residual tonal signal, magnitude spectrum encoding vectors, or a combination of time-domain waveforms and magnitude spectrum encoding vectors.

Abstract: Prior to deriving precise evaluation values for evaluating errors between synthetic signal vectors and a target signal vector, simple evaluation values are derived. Based on the simple evaluation values, a given number of high-ranking candidates are preliminarily selected and then the precise evaluation values are derived with respect to the preliminarily selected candidates. For the preliminary selection of the candidates, the simple evaluation values are divided into as many groups as the number of the candidates to be preliminarily selected. Then, the simple evaluation values are mutually compared in each group to pick up the optimum value in each group.

Abstract: First and second codeword are selected from respective first and second codebooks having an equal number of codewords and wherein the first and second codewords represent unequal numbers of elements of respective first and second sub-vectors. A codebook populating method for a split vector quantizer relies on comparing centroid calculations for the first codebook. The calculations are performed on eligible pairs of codewords. Eligible codewords are limited to those which satisfy and ordered property based on Line Spectrum Frequencies (LSF). The results of the centroid pair codeword calculations are used to populate the second codebook.

Abstract: A high-speed search method in a speech encoder using an order character of LSP (Line Spectrum Pair) counts in a LSP count quantizer using SVQ (Split Vector Quantization) used in a low-speed transmission speech encoder, includes the steps of rearranging a codebook according to an element value of a reference row for determining a range of code vectors to be searched; and determining a search range by using an order character between a given target vector and an arranged code vector to obtain an optimal code vector.

Abstract: An automatic recognition system and method divides observation vectors into subvectors and determines a quantization index for the subvectors. Subvector indices can then be transmitted or otherwise stored and used to perform recognition. In a further embodiment, recognition probabilities are determined for subvectors separately and these probabilities are combined to generate probabilities for the observed vectors. An automatic system for assigning bits to subvector indices can be used to improve recognition.

Abstract: A method and apparatus for compressing and decompressing an audio signal. The apparatus comprises an input for receiving an audio signal derived from a spoken utterance, the audio signal being contained into a plurality of successive data frames. A data frame holding a certain portion the audio signal is processed to generate a feature vector including a plurality of discrete elements characterizing at least in part the portion of the audio signal encompassed by the frame, the elements being organized in a certain sequence. The apparatus makes use of a compressor unit having a grouping processor for grouping elements of the feature vector into a plurality of sub-vectors on the basis of a certain grouping scheme, at least one of the sub-vectors including a plurality of elements from the feature vector, the plurality of elements being out of sequence relative to the certain sequence. The plurality of sub-vectors are then quantized by applying a vector quantization method.

Abstract: A communication apparatus making use of speech vector coding scheme stores a plurality of predetermined code vectors and registered keyword data. A speech encoder codes an input speech keyword to produce coded keyword data by referring to the predetermined code vectors, after detecting a matching level between the registered keyword data and the coded keyword data, it is determined whether the coded keyword data is true or false by comparing the matching level with a predetermined criterion.

Abstract: This invention relates to a method for voice signal coding and/or decoding. According to this method, a voice signal analysis for determining the prediction parameters is carried out from a digital voice signal. An excitation signal component (ENtp) is determined from an adaptive code book (3) built from a delayed integral excitation signal (Ev). Further, a multipulse component of the excitation signal is determined (4) by minimising the effect of the weighting filtered difference between the signal resulting from the respective excitation signal and the input voice signal.

Abstract: A speech communication system using a code excited linear prediction speech decoder. The decoder using a first codebook containing a first digital value sequence selected from the set of binary values {0, 1}. The decoder also using a second codebook containing a second digital value sequence having values selected from the set of binary values {−1, 0}. The first digital value sequence and the second digital value sequence are combined to become a third digital value sequence having a set of ternary values from the set of {−1, 0, 1}.

Abstract: In methods and apparatus for encoding a gain parameter in a generalized linear predictive analysis-by-synthesis (GLPAS) coder, a subframe gain parameter is determined for each of a plurality of successive subframes of a frame, and a quantized frame gain parameter is determined for each frame using a delayed decision quantizer operating on the subframe gain parameters. The subframe gain parameters may be treated as components of a gain vector and the gain vector may be vector quantized to determine the quantized frame gain parameter. Encoder parameters are efficiently aligned with decoder parameters to ensure proper end-to-end operation. Alternatively, tree quantization or trellis quantization may be applied to the subframe gain parameters to determine the quantized frame gain parameter. The methods and apparatus are particularly applicable to low bit rate speech coding.

Abstract: The spectral or pitch parameters of a speech signal are quantized, and impulse responses thereof are predicted by using a filter. An orthogonal transform is made of the speech signal, or a signal derived therefrom, or of the impulse responses or signals derived therefrom. The result of the orthogonal transform is entirely or partly quantized to obtain a plurality of pulses. More preferably, these pulses are retrieved recurrently by also using codevectors retrieved from a codebook or collectively quantizing their senses or amplitudes. This method optimizes speech signal coding.

Abstract: In the novel speech coding method and device, speech signals are coded by a combination of speech parameters and excitation signals. The speech parameters or excitation signals are described with vectors. The vectors are formed by superposing at least two tracks, wherein at least one track has at least two vector elements different from zero. The algebraic signs of the vector elements that differ from zero are coded independently of one another and independently of the positions of the vector elements that differ from zero.

Abstract: An audio encoder 3 divides on a time axis an input audio signal into predetermined coding units and executes coding to each of the coding units so as to output a plurality of types of audio coded parameters. A cyclic redundancy check (CRC) code calculation block 5 selects important bits relative to human hearing from the audio coded parameters of the plurality of types from the audio encoder 3, and creates a CRC check code from the important bits. A convolution encoder 6 executes a convolution coding to the CRC check code and the important bits from the CRC code calculation block.

Abstract: A differential pulse-code modulation coder obtains an improved signal-to-noise ratio, with only a small increase in bit rate, by repetitive coding. In one aspect, the coder divides the input signal into frames, codes each frame repeatedly using different prediction coefficients or different quantizing step functions, and selects the coefficient or step function that produces the least quantization error. In another aspect, the coder repeats the coding of individual samples located in the outermost steps of the quantizing step function.

Abstract: A first CELP coding circuit receiving a signal obtained by down-sampling of an input signal by a down-sampling circuit, outputs a part of coded output to a second CELP coding circuit. The second CELP coding circuit encodes the input signal on the basis of the coded output of the first CELP coding circuit. A multiplexer outputs the coded outputs of the first and second CELP coding circuits in a form of a bit stream. A demultiplexer outputs the coded output of the first CELP coding circuit from the bit stream to a first CELP decoding circuit when a control signal is low bit rate, and extracts a part of the output of the first CELP coding circuit and the output of the second CELP coding circuit to output to a second CELP decoding circuit to output via a switch circuit when the control signal is high bit rate.

Abstract: Speech is encoded into a frame of bits. A speech signal is digitized into a sequence of digital speech samples that are then divided into a sequence of subframes. A set of model parameters is estimated for each subframe. The model parameters include a set of voicing metrics that represent voicing information for the subframe. Two or more subframes from the sequence of subframes are designated as corresponding to a frame. The voicing metrics from the subframes within the frame are jointly quantized. The joint quantization includes forming predicted voicing information from the quantized voicing information from the previous frame, computing the residual parameters as the difference between the voicing information and the predicted voicing information, combining the residual parameters from both of the subframes within the frame, and quantizing the combined residual parameters into a set of encoded voicing information bits which are included in the frame of bits.

Abstract: System efficiently communicates a perceptually encoded speech spectrum signal from a transmitter to a receiver. The transmitter includes a speech analyzer which accepts a speech signal input and generates a parameterized speech signal. The transmitter also includes a vector quantizer for generating the perceptually encoded speech spectrum signal from the parameterized speech signal. The receiver decodes the perceptually encoded speech spectrum signal to produce decoded spectral parameters to further produce a synthetic speech output. The vector quantizer performs a method for partitioning a vector quantizer (VQ) codebook to produce perceptually organized sub-codebooks. The vector quantizer performs a second method for quantizing a vector based on the perceptually organized sub-codebooks. The second method identifies a vector, from one of the perceptually organized sub-codebooks, to perceptually model the speech signal input.

Abstract: Auxiliary multi-pulse setting circuit 130 set candidates of pulse positions so that the pulse positions to which no pulse is located are selected in auxiliary multi-pulse searching circuit 131 prior to the pulse positions at which pulses have already been encoded in multi-pulse searching circuit 110. Auxiliary multi-pulse searching circuit 131 generates an auxiliary multi-pulse signal according to the candidates of pulse positions set in auxiliary multi-pulse setting circuit 130 and encodes the auxiliary multi-pulse signal so that difference between the reproduced audio signal which is obtained by driving a linear predictive synthesis filter with the auxiliary multi-pulse signal and an input audio signal is minimized similarly to multi-pulse searching circuit 110.

Abstract: Two codebooks each consisting of a filter memory are used for vector quantizing of a speech sample. Fixed excitation vectors and pitch parameters of a prediction filter are entered in the respective codebooks, which are actualized in time intervals. To improve the speech quality, respectively two vectors from the adaptive codebook which are best in respect to an error criterion are linked with all vectors of the fixed codebook. The value which best matches an original speech scanned value is selected from the linkages. The entries in the first codebook are advantageously thinned out by suppressing vector components taken from sum bits of two frame sections into which the speech sample is divided until the processing work is no more than the processing work with only one selected best vector from the second codebook.

Abstract: Speech is encoded into a frame of bits. A speech signal is digitized into a sequence of digital speech samples that are then divided into a sequence of subframes. A set of model parameters is estimated for each subframe. The model parameters include a set of spectral magnitude parameters that represent spectral information for the subframe. Two or more consecutive subframes from the sequence of subframes may be combined into a frame. The spectral magnitude parameters from both of the subframes within the frame may be jointly quantized.

Abstract: A family of low-complexity, high quality CELP speech coders are described which use two new techniques: Backward and Inverse Filtered Target (BIFT) for fixed codebook excitation search; and Tree-Structured Multitap adaptive codebook search. Incorporation of these new techniques resulted in very low complexity CELP coders at less than 16 Kb/s. The three coefficients for linear combination of the adaptive codebook are chosen from a tree-structured tap codebook. The best tap index in the primary codebook points to a secondary codebook where the search is further conducted. This procedure may be repeated many times, wherein each subsequent tap codebook points to yet another subsequent tap codebook, which points to yet another subsequent tap codebook, etc. A fixed ternary excitation codebook using a new technique called Backward and Inverse Filtered Target matching (BIFT), is used to encode the portion of the target signal that is left behind after the adaptive codebook contribution has been subtracted.