Sony Corporation

Other

Naoto Iwahashi

Kenzo Akagiri

Masayuki Nishiguchi

Yoshihito Fujiwara

Makoto Akune

Hitoshi Honda

Hongchang Bao

Takushi Kunihiro

Mitsuo Higashiyama

Jun Nakai

Keiichi Yamada

Ralf Kompe

Silke Goronzy

Peter Buchner

Tomoko Umezawa

Hiromi Setoguchi

It is intended to increase the recognition rate in speech recognition and image recognition. An observation vector as input data, which represents a certain point in the observation vector space, is mapped to a distribution having a spread in the feature vector space, and a feature distribution parameter representing the distribution is determined. Pattern recognition of the input data is performed based on the feature distribution parameter.

Feature extraction apparatus and method and pattern recognition apparatus and method

A word is input, along with a concept notation function that is a function for expressing what are described by that word, from a word set output unit to an optimal word string output unit. The optimal word string output unit computes a similarity between a word from the word set output unit and a registered word registered in a dictionary. The optimal word string output unit computes, even when a word not registered in the dictionary is input, a similarity between that word and a word registered in the dictionary.

Information processing device information processing method, and recording medium

A feature extraction and pattern recognition system in which an observation vector forming input data, which represents a certain point in the observation vector space, is mapped to a distribution having a spread in the feature vector space, and a feature distribution parameter representing the distribution is determined. Pattern recognition of the input data is performed based on the feature distribution parameter.

A language learning apparatus includes a syntax-structure analyzing unit and a syntax-rule learning unit. The syntax-structure analyzing unit is provided with a speech recognition portion for recognizing a word string from input speech data O based on speech pattern information of a vocabulary-information storage unit. A semantic analyzer detects image patterns from input moving picture data V by using the information stored in the vocabulary-information storage unit, and analyzes the relevance of the detected image patterns and outputs a concept representation using the words corresponding to the individual image patterns. A relevance analyzer verifies the word string extracted by the speech recognition portion against the semantic representation extracted by the semantic analyzer so as to determine the syntax structure of the word string. Then, the syntax-rule learning unit learns the syntax rule by using the determined syntax structure.

Language learning apparatus and method therefor

To prevent adaptation to misrecognized words in unsupervised or on-line automatic speech recognition systems confidence measures are used or the user reaction is interpreted to decide whether a recognized phoneme, several phonemes, a word, several words or a whole utterance should be used for adaptation of the speaker independent model set to a speaker adapted model set or not and, in case an adaptation is executed, how strong the adaptation with this recognized utterance or part of this recognized utterance should be performed. Furtheron, a verification of the speaker adaptation performance is proposed to secure that the recognition rate never decreases (significantly), but only increases or stays at the same level.

Semi-supervised speaker adaptation

A discriminant or identification function is used for pattern recognition in which the highest performance can be offered when adaptation is made. Learning is carried out while a discriminant or identification function is adapted to a learning sample. For example, a standard pattern of the character 'A' used as an identification function is learned such that when the character 'A' slanting in the right or left direction is input, the standard pattern of the character 'A' is rotated (adapted) in accordance with the slanting of the input learning sample.

Identification-function calculator, identification-function calculating

A parameter converting method and a speech synthesizing method for synthesizing speech with a voice quality similar to that of an input voice includes a parameter converting function constituted of a weighting function for setting weighting coefficients on an input sound spectrum parameter space and a plurality of sub-converting functions. Conversion outputs of the respective sub-converting functions are given weighting coefficients such that the sum of the weighted conversion outputs is used as the parameter converting function to convert M sound spectrum parameter to a single sound spectrum parameter. In this way, the freedom of adaptation with respect to the parameter converting function can be more properly set, so that the parameter converting function can provide an accuracy in accordance with an amount of speech data inputted for learning. It is therefore possible to provide a sound spectrum parameter for generating a voice quality much closer to that of an input voice.

Method for processing speech signal using sub-converting functions and a

A method for speech analysis applicable to a speech analysis/synthesis system employed for producing a synthetic speech. Voiced and unvoiced segments of input speech signals X(n) are discriminated. An amplitude information A(.omega.) and a phase information P.sub.X (.omega.) are extracted from the voiced segments of the input speech signals. A pitch period is detected from the voiced segments of the input speech signals. A pulse train S(n) as a sound source information is generated so that its period corresponds on the time scale to the detected pitch period of the input speech signals. A phase information P.sub.S (.omega.) is extracted from the pulse train S(n). A difference P(.omega.) between the phase information P.sub.S (.omega.) of the pulse train S(n) and the phase information P.sub.X (.omega.) of the input speech signal is found and is supplied as the phase information of the desired one-pitch period within the input speech signals.

Speech analysis method and device for suppyling data to synthesize

A decoder for reproducing information from an encoded signal that has been frequency-domain transformed by a transform conversion circuit in response to audio signal samples grouped into time-domain blocks based on selected block time lengths, includes a reverse conversion transformation circuit that groups the frequency-domain transfer coefficients into transform blocks based on the block time length information used in the encoding operation and applies the inverse discrete transform function to generate time-domain signal sample blocks therefrom.

Signal decoding apparatus

A system for encoding audio signals, transmitting or recording the encoded audio signals, and then decoding the transmitted audio signal includes a circuit for determining the block length of the input audio signals based upon detected values of predetermined portions of the input signal and then performing an orthogonal conversion of the determined block length signal, in which the signal is orthogonally transformed from a time-domain signal to a frequency-domain signal. Following quantization, the converted and encoded signal is transmitted and upon performing reverse quantization, a reverse orthogonal converting circuit converts the signal back to the time-domain in keeping with the signal providing information concerning the various block lengths that were determined during encoding. The converted time-domain signals are then connected into a smooth waveform to form output audio signals.

Signal encoding and signal decoding apparatus

A signal encoding apparatus for cutting out blocks of an input signal waveform at a predetermined time interval and converting the in-block signal by mutually independent conversion axes for encoding, wherein the waveform cutting block length along the time axis is changed according to the input signal.

Signal encoding apparatus

An apparatus for predictive coding processing a digital signal comprises a selective predictive filter, responsive to an input signal for outputting a predictive residual signal, an adaptive predictive filter, responsive to the predictive residual signal, for outputting an optimal predictive residual signal, and a re-quantizing unit means for re-quantizing and outputting the optimal predictive residual signal. The positions of the selective and adaptive predictive filters may be interchanged and the processing apparatus may be either feed forward or feed back type.

Growing community of inventors

Kanagawa, Japan

Naoto Iwahashi