In a language translation system, a method for providing a guide for a user. In one embodiment, the method comprises the steps of receiving (302) an input that is representative of at least one word in a source language, and generating (1402) at least one recognition hypothesis in the source language in response to the input. The method further comprises the steps of selecting (406) a best hypothesis from the at least one recognition hypothesis in the source language, presenting (408) the best hypothesis in the source language to a user, and presenting (408) alternatives to a portion of the best hypothesis in the source language to the user. The method further includes receiving (410) an indication of a choice of one of the alternatives from the user, and presenting (306) a revised version of the best hypothesis including the alternative chosen to the user.

A method and apparatus for interactive source language expression recognition and alternative hypothesis presentation and selection

A continuous speech recognition and voice response system provides a natural sounding and effective interactive, speech-driven dialogue from a data processing system. A concatenation of words into phrases and sentences improves recognition and mimic natural language processing. The system uses speaker-independent, continuous-speech to initiate the playback of audio files. The system employs high-speed context switching to modify the active vocabulary and applies high-speed context switching to modify or activate Audio WAV voice response files. The system uses dialogue history to activate selected context, Baukus-Naur Form (BNF) grammars and WAV files and provides phrase or sentence long dialogue prompts to improve accuracy. The system also provides audio prompts to improve accuracy and provides speech-activated buttons to navigate between menus.

Continuous speech recognition and voice response system and method to enable conversational dialogues with microprocessors

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. The eigenvoice technique may be applied to MLLR transformation matrices or the like; Bayesian estimation performed in eigenspace uses prior knowledge about speaker space density to refine the estimate about the location of a new speaker in eigenspace.

Dimensionality reduction for speaker normalization and speaker and environment adaptation using eigenvoice techniques

A process and apparatus for measuring the resemblance between sound samples incorporating a learning phase and a test phase, each having a digital acquisition and preprocessing stage, an order p acoustic analysis stage and a p×p covariance matrix calculating stage with respect to the resulting vector signal, the learning phase also having a reference X covariance matrix inversion stage and a stage of storing the thus obtained matrix X-1 in a dictionary, the test phase incorporating a stage of multiplying the covariance matrix of test Y with the reference matrix X-1, a stage of extracting all or part of the p eigenvalues λk of said matrix product and a calculating stage using a family of functions f, called generalized sphericity functions, so as to obtain a digital value measuring the resemblance between the considered test sample and the reference samples of the dictionary. Alternative implementations are also described, which advantageously allow the same result to be obtained without requiring the explicit calculation of eigenvalues.

Process for measuring the resemblance between sound samples and apparatus for performing this process

A telephony channel simulation process is disclosed for training a speech recognizer to respond to speech obtained from telephone systems. An input speech data set is provided to a speech recognition training processor, whose bandwidth is higher than a telephone bandwidth. The process performs a series of alterations to the input speech data set to obtain a modified speech data set. The modified speech data set enables the speech recognition processor to perform speech recognition on voice signals from a telephone system.

Telephony channel simulator for speech recognition application

In a speech recognition system, the beginning of speech versus non-speech (a cough or noise) is distinguished by reverting to a non-speech decision process whenever the liklihood cost of template (vocabulary) patterns, including silence, is worse than a predetermined threshold, established by a Joker Word which represents a non-vocabulary word score and path in the grammar graph.

Speech recognition method having noise immunity

A speech recognition method and apparatus employ a speech processing circuitry for repetitively deriving from a speech input, at a frame repetition rate, a plurality of acoustic parameters. The acoustic parameters represent the speech input signal for a frame time. A plurality of template matching and cost processing circuitries are connected to a system bus, along with the speech processing circuitry, for determining, or identifying, the speech units in the input speech, by comparing the acoustic parameters with stored template patterns. The apparatus can be expanded by adding more template matching and cost processing circuitry to the bus thereby increasing the speech recognition capacity of the apparatus. Template pattern generation is advantageously aided by using a "joker" word to specify the time boundaries of utterances spoken in isolation, by finding the beginning and ending of an utterance surrounded by silence.

Speech recognition training method

A speech recognition method and apparatus employ a speech processing circuity (26) for repetitively deriving from a speech imput (100), at a frame repetition rate, a plurality of acoustic parameters. The acoustic parameters represent the speech input signal for a frame time. A plurality of template matching and cost processing circuitries (28, 30) are connected to a system bus (24), along with the speech processing circuity, for determining, or identifying, the speech units in the input speech, by comparing the acoustic parameters with stored template patterns. The apparatus can be expanded by adding more template matching and cost processing circuity to the bus thereby increasing the speech recognition capacity of the apparatus. The speech processing circuity establishes overlapping time durations for generating the acoustic parameters and further employs a sinc-Kaiser smoothing function in combination with a folding technique (113) for providing a discrete Fourier transform (112). The Fourier spectra are transformed using a principal component analysis (122) which optimizes the across class variance. The template matching and cost processing circuitries (28, 30) provide distributed processing, on demand, of the acoustic parameters for generating through a dynamic programming technique the recognition decision. Grammar and word model syntax structures reduce the computational load. Template pattern generation is aided by using a "joker" word to specify the time boundaries of utterances spoken in isolation.

Chin-hui Lee

Papers

Speech recognition method having noise immunity

Speech recognition training method

Speech recognition methods and apparatus