Abstract: Square sum calculator 603 calculates a square sum of evolution in smoothed quantized LSP parameter for each order. A first dynamic parameter is thereby obtained. Square sum calculator 605 calculates a square sum using a square value of each order. The square sum is a second dynamic parameter. Maximum value calculator 606 selects a maximum value from among square values for each order. The maximum value is a third dynamic parameter. The first to third dynamic parameters are output to mode determiner 607, which determines a speech mode by judging the parameters with respective thresholds to output mode information.

Abstract: A dispersed vector generator used for a speech encoder or a speech decoder includes a pulse vector provider that provides a pulse vector having a signed unit pulse on one element of a vector axis. A dispersion pattern determiner determines a dispersion pattern of a set of waveforms defined before a start of encoding or decoding. A dispersed vector generator convolutes the pulse vector and the determined dispersion pattern to generate a dispersed vector. A length of the waveforms is shorter than a length of a sub-frame.

Abstract: An apparatus sets a pitch cycle search object in pitch cycle search processing for searching for a pitch cycle included in a linear predictive residual on a per subframe basis. A pitch cycle indicator of the apparatus sequentially output spitch cycle candidates within a predetermined pitch cycle search range at integral accuracy. A memory stores an integral component of a pitch cycle selected in pitch cycle search processing of a previous subframe. An adaptive sound source vector generator sets, as the pitch cycle search object in pitch cycle search processing in a processing subframe section, a group of candidates comprising a group of integral-accuracy pitch cycle candidates output from the pitch cycle indicator and a group of fractional-accuracy pitch cycle search candidates that cover a pitch cycle near an integral component of the pitch cycle read from the previous subframe integral pitch cycle memory using fractional accuracy.

Abstract: Speech is coded using an orthogonal search by calculating a search reference value. An adaptive codevector representing a pitch component is generated. A random codevector representing a random component is also generated. The orthogonal search further includes generating a synthetic speech signal by a synthesis filter being excited by the adaptive codevector and the random codevector. A distortion between the input speech signal and the synthetic speech signal is calculated. One random codevector that minimizes the distortion is selected.

Abstract: This invention decoded encoded speech using alternative parameters upon detection of a lost packet. Upon detection of a first good packet following packet loss, this invention uses second alternative parameters intermediate between the default parameters and the alternative parameters for a predetermined interval. Thereafter the invention reverts to the default parameters. This minimizes glitches in the decoded speech upon packet loss. This invention is suitable for use in decoding speech data encoded in the CCITT Recommendation G.726 ADPCM based speech coding standard.

Abstract: A fixed codebook searching apparatus which slightly suppresses an increase in the operation amount, even if the filter applied to the excitation pulse has the characteristic that it cannot be represented by a lower triangular matrix and realizes a quasi-optimal fixed codebook search. This fixed codebook searching apparatus is provided with an algebraic codebook (101) that generates a pulse excitation vector; a convolution operation section (151) that convolutes an impulse response of an auditory weighted synthesis filter into an impulse response vector that has a value at negative times, to generate a second impulse response vector that has a value at second negative times; a matrix generating section (152) that generates a Toeplitz-type convolution matrix by means of the second impulse response vector; and a convolution operation section (153) that convolutes the matrix generated by matrix generating section (152) into the pulse excitation vector generated by algebraic codebook (101).

Abstract: An encoding device (200) includes an MDCT unit (202) that transforms an input signal in a time domain into a frequency spectrum including a lower frequency spectrum, a BWE encoding unit (204) that generates extension data which specifies a higher frequency spectrum at a higher frequency than the lower frequency spectrum, and an encoded data stream generating unit (205) that encodes to output the lower frequency spectrum obtained by the MDCT unit (202) and the extension data obtained by the BWE encoding unit (204). The BWE encoding unit (204) generates as the extension data (i) a first parameter which specifies a lower subband which is to be copied as the higher frequency spectrum from among a plurality of the lower subbands which form the lower frequency spectrum obtained by the MDCT unit (202) and (ii) a second parameter which specifies a gain of the lower subband after being copied.

Abstract: To alleviate problems of signal aliasing and to reduce complexity, Linear Predictive Coefficients (LPCS) are calculated from samples of audio signals and Line Spectral Frequency (LSF) vectors are extracted from the LPCs with a rate higher than a desired vector rate, the LSF vectors comprising values of different LSF parameters. Next, an LSF track is formed for at least one of the LSF parameters. At least one of the formed LSF tracks is then low pass filtered. Finally, decimated LSF vectors are reconstructed from the low pass filtered LSF tracks, the decimated number corresponding to the desired vector rate. The invention equally relates to a corresponding computer program, to corresponding devices and to a corresponding communication network.

Abstract: A method and apparatus for implementing fixed codebooks as a common module are provided. In the method of implementing fixed codebooks of a plurality of speech codecs as a common module, it is possible to include only a part excluding fixed codebooks in a communication terminal or communication system, support various speech codecs without using a chip with high price and high performance, and reduce a memory space that is occupied by the speech codecs by generating a track of a fixed codebook corresponding to a speech codec based on information on the speech codec among the plurality of speech codecs and selecting a codebook vector corresponding to a target signal among codebook vectors constructed with combinations of pulses represented by the generated track. In addition, it is possible to reduce processing complexity as compared with a case of embodying the common fixed codebook module in software by embodying the common fixed codebook module in hardware.

Abstract: A non-intrusive signal quality assessment apparatus includes a feature vector calculator that determines parameters representing frames of a signal and extracts a collection of per-frame feature vectors (?;(n)) representing structural information of the signal from the parameters. A frame selector preferably selects only frames (?\with a feature vector (?;(n)) lying within a predetermined multi-dimensional window (?). Means determine a global feature set (?) over the collection of feature vectors (?;(n)) from statistical moments of selected feature vector components ((1?,01, . . . O11). A quality predictor predicts a signal quality measure (Qj from the global feature set (?)).

Abstract: Each of the M basic vectors in a noise code book 260 is multiplied by a factor ±1 in a sign adder 270 and combined in an adder 280 to create 2M noise signed vectors. The characteristic of the binary Gray code is utilized as follows. A change ?Gu obtained between a noise signed vector based on a signed word i of the binary Gray code and a noise sign vector based on a sign word u adjacent to the sign word i and different from the sign word i only in a predetermined bit position v is used in such a manner that a sign word u? which is next to reverse the bit position v on the Gray code sequence can express a change ?Gu? from the noise signed vector by utilizing the fact that the sign word u? differs from the sign word u only in one bit position w excluding the bit position V. Thus, calculation is simplified, increasing the vector search speed.

Abstract: There is provided a method of using an adaptive tilt compensation by a speech decoder. The method comprises receiving a bit stream including a plurality of parameters representative of a speech signal; identifying an adaptive code vector and a fixed code vector using the plurality of parameters; scaling the adaptive code vector and the fixed code vector to generate a scaled adaptive code vector and a scaled fixed code vector; summing the scaled adaptive code vector and the scaled fixed code vector to generate a synthesized output; calculating a first reflection coefficient based on the plurality of parameters representative of the speech signal; multiplying the first reflection coefficient by a factor to generate a tilt factor; and applying the tilt factor to the synthesized output based on an encoding bit rate.

Abstract: Methods and arrangements for generating a voice model in speech processing. Upon accepting at least two input vectors with spectral features, vectors of ranks are created via ranking values of the spectral features of each input vector, ordered vectors are created via arranging the values of each input vector according to rank, and a vector of ordered average values is created via determining the average of corresponding values of the ordered vectors. Thence, a vector of ordered average ranks is created via determining the sum or average of the vectors of ranks, a vector of ordered ranks is created via ranking the values of the ordered average ranks and a spectral feature vector is created via employing the rank order represented by the vector of ordered ranks to reorder the vector of ordered average ranks.

Abstract: The input signal can be quickly and accurately classified and a descriptor can be generated according to the result of classification. Then, the input signal can be retrieved on the basis of the result of classification or the descriptor. A signal processing apparatus comprises a time block splitting section 3 for splitting an audio signal into blocks that are typically 1 second long, a feature extracting section 4 for extracting a characteristic quantity of 18 degrees on the signal attribute from the audio signal in each block and a vector quantizing section 5 for carrying out an operation of categorical classification for the audio signal of each block by means of a vector quantization technique that uses a VQ code book 8 and a characteristic vector formed from the characteristic quantity of 18 degrees. The vector quantizing section 5 outputs a classification label obtained as a result of the categorical classification and a descriptor indicating the reliability of the label.

Abstract: A frame type for a current SBR frame is determined according to a type of end border of a previous frame, as well as presence of a transient in the current SBR frame. A start border is determined according to the end border of the previous SBR frame. For a FIXFIX frame, a low time-resolution setting is used. For a FIXVAR or a VARVAR frame, a search for intermediate borders is conducted in the region between the transient and maximum allowed end border location. The end border is also determined at this stage. If there is excess capacity for more borders, another search is conducted in the region between the transient and the start border. For a VARFIX frame, only one search needs to be conducted, in the whole region partitioned by a variable start border and a fixed end border. All of the above are accomplished with two Forward Search operations and one Backward Search operation.

Abstract: A speech encoder includes an adaptive codebook that generates an adaptive codevector representing a pitch component, a random codebook that generates a random codevector representing a random component, and a synthesis filter that generates a synthetic speech signal by being excited by the adaptive codevector and the random codevector. The random codebook includes an input vector provider configured to provide an input vector, and an excitation vector generator configured to generate an excitation vector as the random codevector by dispersing the input vector by using a fixed pattern. A length of the fixed pattern is shorter than a length of a sub-frame.

Abstract: Provided are a vocal-cord recognition apparatus and a method thereof. The vocal-cord signal recognition apparatus includes a vocal-cord signal extracting unit for analyzing a feature of a vocal-cord signal inputted through a throat microphone, and extracting a vocal-cord feature vector from the vocal-cord signal using the analyzing data; and a vocal-cord signal recognition unit for recognizing the vocal-cord signal by extracting the feature of the vocal-cord signal using the vocal-cord signal feature vector extracted at the vocal-cord signal extracting means.

Abstract: An apparatus for providing voice conversion using temporal dynamic features includes a feature extractor and a transformation element. The feature extractor may be configured to extract dynamic feature vectors from source speech. The transformation element may be in communication with the feature extractor and configured to apply a first conversion function to a signal including the extracted dynamic feature vectors to produce converted dynamic feature vectors. The first conversion function may have been trained using at least dynamic feature data associated with training source speech and training target speech. The transformation element may be further configured to produce converted speech based on an output of applying the first conversion function.

Abstract: Methods, encoders, and digital systems are provided for predictive encoding of speech parameters in which an input frame is encoded by quantizing a parameter vector of the input frame with a strongly-predictive codebook and a weakly-predictive codebook to obtain a strongly-predictive distortion and a weakly-predictive distortion, adjusting a correlation indicator based on a relative correlation of the input frame to a previous frame, wherein the correlation indicator is indicative of the strength of the correlation of previously encoded frames, and encoding the input frame with the weakly-predictive codebook unless the correlation indicator has reached a correlation threshold.

Abstract: A characteristic thumbprint is extracted from a data signal, the thumbprint based on statistics relating to the data signal. The data signal can be compared indirectly by matching this thumbprint against one or more reference thumbprints. The data signal may be any type of signal, including streaming digitized audio or obtained from static files. A database for containing a number of such thumbprints and methods for searching the database are also described.

Abstract: A method and apparatus for performing multiple descriptive source coding in which a plurality of homogeneous encoders are advantageously employed in combination with a corresponding plurality of advantageously substantially identical decoders. In particular, diversity is provided to the multiple encoders by modifying the quantization process in at least one of the encoders such that the modified quantization process is based at least on a quantization error resulting from the quantization process of another one of the encoders. In this manner, diversity among the multiple bit streams is obtained, and in particular, the quality of a reconstructed signal based on a combination of multiple decoded bit streams at the receiver is advantageously superior to that based on any one of the decoded bit streams. In accordance with a first illustrative embodiment of the present invention, two Pulse Code Modulation (PCM) coders are employed.

Abstract: A decoding apparatus is provided. The decoding apparatus has a first decoding part for decoding a code word obtained by encoding an input signal using a Code-Excited Linear Prediction encoding method. A second decoding part decodes a code word obtained by encoding a signal with an encoding method other than the Code-Excited Linear Prediction encoding method. A rising-transition detection and notification part has a detection part that detects the existence of a rising-transition of amplitude of the input signal based on time variation of a gain of excitation vectors obtained by the first decoding part, and a notification part that notifies the second decoding part that the rising-transition of the amplitude exists.

Abstract: A method and apparatus for language independent voice searching in a mobile communication device is disclosed. The method may include receiving a search query from a user of the mobile communication device, converting speech parts in the search query into linguistic representations which covers at least one languages, generating a search phoneme lattice based on the linguistic representations, extracting query features from the search phoneme lattice, generating query feature vectors based on the extracted features, performing a coarse search using the query feature vectors and the indexing feature vectors from the indexing database, performing a fine search using the results of the coarse search and the indexing phoneme lattices stored in the indexing database, and outputting the fine search results to a dialog manager.

Abstract: The present invention carries out pre-selection on many LPC codevectors stored in an LSF codebook 101 using a weighted Euclidean distortion as a measure and carries out a full-code selection on the LPC codevectors left after the pre-selection using an amount of distortion in a spectral space as a measure. This makes it possible to improve the quantization performance of the LPC parameter vector quantizer and improve the quality of synthesized speech of the speech coder/decoder.

Abstract: According to the invention, quantization encoding is conducted using the probability density function of the source, enabling fixed, variable and adaptive rate encoding. To achieve adaptive encoding, an update is conducted with a new observation of the data source, preferably with each new observation of the data source. The current probability density function of the source is then estimated to produce codepoints to vector quantize the observation of the data source.

Abstract: A high-speed search method in a speech encoder using an order character of LSP (Line Spectrum Pair) parameters in an LSP parameter 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. The method gives effects of reducing computational complexity required to search the codebook without signal distortion in quantizing the LSP parameters of the speech encoder using SVQ, and reducing computational complexity without loss of tone quality in G.729 fixed codebook search by performing candidate selection and search on the basis of the correlation value size of the pulse position index.

Abstract: CELP-based speech encoder that performs encoding by decomposing one frame into a plurality of subframes, includes an LPC synthesizer that obtains synthesized speech by filtering an adaptive excitation vector and a stochastic excitation vector stored in an adaptive codebook and in an stochastic codebook using LPC coefficients obtained from input speech. A gain calculator calculates gains of the adaptive excitation vector and the stochastic excitation vector. A parameter coder performs vector quantization of the adaptive excitation vector and the stochastic excitation vector obtained by comparing distortions between the input speech and the synthesized speech. A pitch analyzer performs pitch analyses of a plurality of subframes in the frame respectively, before performing an adaptive codebook search for the first subframe, calculating correlation values and finding a value most approximate to the pitch period using the correlation values.

Abstract: A method for estimating the performance of a statistical classifier. The method includes inputting a first set of business data in a first format from a real business process and storing the first set of business data in the first format into memory. The method applying a statistical classifier to the first set of business data and recording its classification decisions and obtaining a labeling that contains the correct decision for each data item. The method includes computing a weight for each data item that reflects its true frequency and computing a performance measure of the statistical classifier based on the weights that reflect true frequency. The method also displays the performance measure to a user.

Abstract: There is provided a vector conversion device for converting a reference vector used for quantization of an input vector so as to improve a signal quality including audio. In this vector conversion device, a vector quantization unit (902) acquires a number corresponding to a decoded LPC parameter of a narrow band from all the code vectors stored in a code book (903). A vector dequantization unit (904) references the number of the code vector obtained by the vector quantization unit (902) and selects a code vector from the code book (905). A conversion unit (906) performs calculation by using a sampling-adjusted decoded LPC parameter obtained from an up-sampling unit (901) and a code vector obtained from the vector dequantization unit (904), thereby obtaining a decoded LPC parameter of a broad band.

Abstract: The present invention provides, methods, computer-readable media, and apparatuses for tuning and adjusting the computational complexity of algorithm that is executed by a signal encoder. The signal encoder may comprise a speech encoder. When a resource shortage on a computer platform is detected, a degree of the resource shortage and a corresponding complexity adjustment for a speech encoder are determined. The speech encoder is then tuned to adjust the computational complexity of an executed speech processing algorithm. The resource shortage may correspond to a computational capability, audio buffer memory, or battery of a mobile device. A speech process being executed by the mobile device is tuned to adjust the computational demands in accordance with a complexity adjustment. A number of iteration rounds may be adjusted while the speech encoder is executing a speech processing algorithm. The iterations may correspond to an algebraic codebook search.

Abstract: An excitation vector generator is provided for generating an excitation vector. The excitation vector includes a pulse vector generator that has at least one channel for generating pulse vectors. A memory stores at least one type of dispersion pattern for each of the channels. A dispersion pattern is selectively extracted from the memory for each of the channels. A dispersed vector generator generates a dispersed vector for every channel by convolution calculation using the extracted dispersion pattern and the generated pulse vectors. An excitation vector generator generates an excitation vector from the dispersed vectors generated by the dispersion vector generator.

Abstract: The present invention provides methods and apparatus for adjusting an algorithmic time delay of a signal encoder. An input signal is sampled at a predetermined sampling rate. When look-ahead operation is initiated, the algorithmic time delay is increased by the look-ahead time duration. When look-ahead operation is terminated, the algorithmic time delay is decreased by the look-ahead time duration. A set of input signal samples is aligned in accordance with the algorithmic time delay, and an output signal that is representative of the set of signal samples is formed. A first signal segment is added to an input signal waveform when the look-ahead operation is initiated, and a second signal segment is removed from the input signal waveform when the look-ahead operation is terminated. Pointers that point to a beginning of the current frame and to new input signal samples are adjusted when the operational mode changes.

Abstract: Methods, devices, and systems for coding and decoding audio are disclosed. Digital samples of an audio signal are transformed from the time domain to the frequency domain. The resulting transform coefficients are coded with a fast lattice vector quantizer. The quantizer has a high rate quantizer and a low rate quantizer. The high rate quantizer includes a scheme to truncate the lattice. The low rate quantizer includes a table based searching method. The low rate quantizer may also include a table based indexing scheme. The high rate quantizer may further include Huffman coding for the quantization indices of transform coefficients to improve the quantizing/coding efficiency.

Abstract: A method and apparatus for reducing the complexity of linear prediction analysis-by synthesis (LPAS) speech coders. The speech coder includes a multi-tap pitch predictor having various parameters and utilizing an adaptive codebook subdivided into at least a first vector codebook and a second vector codebook. The pitch predictor removes certain redundancies in a subject speech signal and vector quantizes the pitch predictor parameters.

Abstract: A mapping transform unit subjects input audio signals to a mapping transform and generates frequency region signals that take frequency as a variable; a code amount designation unit supplies a preset coding bit rate as a code amount output; a frequency region signal compression encoder, based on the code amount, subjects input frequency region signals to a compression encoding process and generates a bitstream; and a bandwidth-limiting unit executes a bandwidth-limiting processing in which a part of the frequency zone covered by frequency region signals is allotted to an attenuation frequency zone, and in which the value of the frequency region signal is multiplied by an attenuation coefficient having a value less than 1 in the attenuation frequency zone to attenuate the frequency region signal in the attenuation frequency zone, and supplies the frequency region signals that have undergone the bandwidth-limiting processing to the frequency region signal compression encoder.

Abstract: A codebook excited linear prediction coding system providing improved digital speech coding for high quality speech at low bit rates with side-by-side codebooks for segments of the modeled input signal to reduce the complexity of the codebook search. A linear predictive filter responsive to an input signal desired to be modeled is used for identifying a basis vector from a first codebook over predetermined intervals as a subset of the input signal. A long term predictor and a vector quantizer provide synthetic excitation of modeled waveform signal components corresponding to the input signal desired to be modeled from side-by-side codebooks by providing codevectors with concatenated signals identified from the basis vector over the predetermined intervals with respect to the side-by-side codebooks. Once a codevector is identified, the codebook at the next segment is searched and a concatenation of codevectors is provided by selecting up to but not including the current segment.

Abstract: A method and apparatus is disclosed herein for quantizing data using a perceptually relevant search of multiple quantization patterns. In one embodiment, the method comprises performing a perceptually relevant search of multiple quantization patterns in which one of a plurality of prototype patterns and its associated permutation are selected to quantize the target vector, each prototype pattern in the plurality of prototype patterns being capable of directing quantization across the vector; converting the one prototype pattern, the associated permutation and quantization information resulting from both to a plurality of bits by an encoder; and transferring the bits as part of a bit stream.

Abstract: A system and method reduces the effects of the bit-error induced distortion of decoded voice transmission by assigning vectors that are close or similar in Euclidean distance to respective indices that are close in Hamming distance. The system calculates a first distortion sum of the distance error induced by single, double or N bit error possibilities, switches vector assignments and calculates a second distortion sum. If the second sum is less than the first sum the vector swap is maintained.

Abstract: The invention relates to representation of one and multidimensional signal vectors in multiple nonorthogonal domains and design of Vector Quantizers that can be chosen among these representations. There is presented a Vector Quantization technique in multiple nonorthogonal domains for both waveform and model based signal characterization. An iterative codebook accuracy enhancement algorithm, applicable to both waveform and model based Vector Quantization in multiple nonorthogonal domains, which yields further improvement in signal coding performance, is disclosed. Further, Vector Quantization in multiple nonorthogonal domains is applied to speech and exhibits clear performance improvements of reconstruction quality for the same bit rate compared to existing single domain Vector Quantization techniques. The technique disclosed herein can be easily extended to several other one and multidimensional signal classes.

Abstract: When a voice encoding apparatus embeds any data in encoded voice code, the apparatus determines whether data embedding condition is satisfied using a first element code from among element codes constituting the encoded voice code, and a threshold value. If the data embedding condition is satisfied, the apparatus embeds optional data in the encoded voice code by replacing a second element code with the optional data. When a voice decoding apparatus extracts data that has been embedded in encoded voice code, the apparatus determines whether data embedding condition is satisfied using a first element code from among element codes constituting the encoded voice code, and a threshold value. If the data embedding condition is satisfied, the apparatus determines that optional data has been embedded in the second element code portion of the encoded voice code and extracts this embedded data.

Abstract: A method and system for analysis-by-synthesis encoding of an information signal is provide. The encoder (400) can include the steps of generating a first synthetic signal based on a first pitch-related codebook (402), generating a second synthetic signal based on a second pitch-related codebook (404), selecting a codebook configuration parameter based on the reference signal and the first and second synthetic signals, and conveying the codebook configuration for use in reconstructing an estimate of the input signal. The encoder can include an error expression having an error bias (506) and a prediction gain having a prediction gain bias (508) for determining the codebook configuration. The encoder can employ variable length coding and combinatorial subframe coding (600) for efficiently compressing the codebook configuration parameter and codebook related parameters for one or more subframes.

Abstract: The quantizers of delta sigma modulators in the signal processing systems described herein use a reduced set of comparators for quantization by predetermining and maintaining a maximum per cycle deviation d between a loop filter output signal VLF(t) and a predicted quantizer output signal qest. In at least one embodiment, a maximum quantizer level deviation d is defined in terms of a number of quantization levels. Thus, the number of comparators in a quantizer needed to quantize the quantizer input signal Vin(t) is based on the maximum quantizer level deviation d. In addition to using fewer comparators than available quantization output levels N, the quantizers can use an even number of comparators M, in contrast to comparable conventional reduced comparator ADC tracking quantizer designs using M+1 number of comparators, where N and Mare integers and M<N.

Abstract: An audio encoder regulates quality and bitrate with a control strategy. The strategy includes several features. First, an encoder regulates quantization using quality, minimum bit count, and maximum bit count parameters. Second, an encoder regulates quantization using a noise measure that indicates reliability of a complexity measure. Third, an encoder normalizes a control parameter value according to block size for a variable-size block. Fourth, an encoder uses a bit-count control loop de-linked from a quality control loop. Fifth, an encoder addresses non-monotonicity of quality measurement as a function of quantization level when selecting a quantization level. Sixth, an encoder uses particular interpolation rules to find a quantization level in a quality or bit-count control loop. Seventh, an encoder filters a control parameter value to smooth quality. Eighth, an encoder corrects model bias by adjusting a control parameter value in view of current buffer fullness.

Abstract: Linear predictive system with classification of LP residual Fourier coefficients into two or more overlapping classes, and each class has its own vector quantization codebook(s). The use of strong and weak predictors minimizes codebook size by only quantizing the difference between Fourier coefficients of a frame and the Fourier coefficients predicted from a prior frame. The choice of using either a strong or weak predictor adapts to the prior choice of predictor so that a strong predictor following a weak predictor is changed to a weak predictor to insure attenuation of error propagation as arise from frame erasures.

Abstract: An encoding device (100) includes (i) a first encoding unit (132) that encodes spectral data in the lower frequency band represented by a plularity of parameters, out of the spectral data obtained by transforming an audio signal inputted for a fixed time length, (ii) a second quantizing unit (133) that generates sub information representing characteristics of the spectral data in the higher frequency by fewer parameters than those for the lower frequency band, out of the spectral data obtained by the transformation, (iii) a second encoding unit (134) that encodes the generated sub information, and (iv) a stream output unit (140) that outputs the data encoded by the first encoding unit (132) and the data encoded by the second encoding unit (134).

Abstract: The indexing method comprises forming a set of tracks of pulse positions, restraining the positions of the non-zero-amplitude pulses of the combinations of the codebook in accordance with the set of tracks of pulse positions, and indexing in the codebook each non-zero-amplitude pulse of the combinations at least in relation to the position of the in the corresponding track, the amplitude of the pulse, and the number of pulse positions in said corresponding track. For indexing the position(s) of one and two non-zero amplitude pulse(s) in one track, procedures code—1 pulse and code—2 pulse are respectively used. When the positions of a number X of non-zero-amplitude pulses are located in one track, X?3, subindices of these X pulses are calculated using the procedures code—1 pulse and code—2 pulse, and a global index is calculated by combining these subindices.

Abstract: A method of quantizing a model parameter includes applying the model parameter to a non-linear scaling function to produce a scaled model parameter and quantizing the scaled model parameter to form a quantized model parameter. In further embodiments, likelihoods for multiple frames of input feature vectors are determined for each retrieval of quantized model parameters from memory.

Abstract: In a speech recognition system, a method of transforming speech feature vectors associated with speech data provided to the speech recognition system includes the steps of receiving likelihood of utterance information corresponding to a previous feature vector transformation, estimating one or more transformation parameters based, at least in part, on the likelihood of utterance information corresponding to a previous feature vector transformation, and transforming a current feature vector based on maximum likelihood criteria and/or the estimated transformation parameters, the transformation being performed in a linear spectral domain. The step of estimating the one or more transformation parameters includes the step of estimating convolutional noise Ni? and additive noise Ni? for each ith component of a speech vector corresponding to the speech data provided to the speech recognition system.

Abstract: A method and apparatus for a voice transcoder that converts a bitstream representing frames of data encoded according to a first voice compression standard to a bitstream representing frames of data according to a second voice compression standard using perceptual weighting that uses tuned weighting factors, such that the bitstream of a second voice compression standard to produce a higher quality decoded voice signal than a comparable tandem transcoding solution.

Abstract: An audio encoder regulates quality and bitrate with a control strategy. The strategy includes several features. For example, an encoder selects a quantization level within a range of quantization levels, where the selecting accounts for non-monotonicity of quality measure as a function of quantization level within the range. The encoder then quantizes audio information by the quantization level. Or, an encoder determines first and second quality measures associated with a first and second quantization levels, respectively, then determines a third quantization level within a quantization level range based upon location of a target quality on a trajectory of quality measure as a function of quantization level. The first and second quantization levels define endpoints of the quantization level range, and the first and second quality measures define endpoints of the trajectory. The function relates logarithm of quality measure in proportion to inverse logarithm of quantization level.