Panasonic Corporation

Matsushita Electric Industrial Co., Ltd.

Keiko Morii

Jean-Claude Junqua

Luca Rigazio

David Kryze

Nobuyuki Kunieda

Takeo Kanamori

Mitsuru Endo

Tomohiro Konuma

Maki Yamada

Kazuya Nomura

Koichiro Mizushima

Yoshio Ohno

A hearing aid device includes a plurality of microphones which converts picked sound to audio signals, an inter-signal phase difference calculation unit which calculates a phase difference between a first audio signal and a second audio signal, a hand placement determination unit, a directivity sound pickup unit which generates an output signal, a nonlinear amplification unit which controls the signal level of the output signal acquired from the directivity sound pickup unit, and a hearing aid process control unit which, when the hand placement determination unit determines that a state where a phase difference is equal to or smaller than a first threshold value continues for a first time, controls the directivity sound pickup unit such that at least one of the sensitivity-frequency characteristic and the frequency characteristics of the audio signals becomes nondirective, and controls the nonlinear amplification unit to amplify the signal levels of the acquired audio signals.

Hearing aid device and hearing aid method

The present invention aims at extracting a keyword of conversation without preparations by advanced anticipation of keywords of conversation. A keyword extracting device of the present invention includes an audio input sectionby way of which a speech sound made by a speaker is input; a speech segment determination sectionthat determines a speech segment for each speaker in connection with the input speech sound; a speech recognition sectionthat recognizes a speech sound of the determined speech segment for each speaker; an interrupt detection sectionthat detects a feature of a speech response suggesting presence of a keyword on the basis of a response of another speaker to speech sounds of respective speakers; namely, an interrupt where a preceding speech and a subsequent speech overlap; a keyword extraction sectionthat extracts the keyword from the speech in the speech segment specified on the basis of an interrupt; a keyword search sectionthat performs keyword search by means of the keyword; and a display sectionthat displays a result of keyword search.

Device for extracting keywords in a conversation

Model compression is combined with model compensation. Model compression is needed in embedded ASR to reduce the size and the computational complexity of compressed models. Model-compensation is used to adapt in real-time to changing noise environments. The present invention allows for the design of smaller ASR engines (memory consumption reduced to up to one-sixth) with reduced impact on recognition accuracy and/or robustness to noises.

Block-diagonal covariance joint subspace tying and model compensation for noise robust automatic speech recognition

A noise robustness method operates jointly in a signal domain and a model domain. For example, energy is added in the signal domain for frequency bands where an actual noise level of an incoming signal is lower than a noise level used to train models, thus obtaining a compensated signal. Also, energy is added in the model domain for frequency bands where noise level of the incoming signal or the compensated signal is higher than the noise level used to train the models. Moreover, energy is never removed, thereby avoiding problems of higher sensitivity of energy removal to estimation errors.

Joint signal and model based noise matching noise robustness method for automatic speech recognition

An acoustic analysis unit acoustically analyzes a first utterance by a user. A pattern by-characteristic selection unit selects a trained pattern optimal for the user's utterance from a plurality of trained patterns that are previously classified and stored every characteristic. A speaker adaptation processor determines a spectral frequency distortion coefficient for correcting difference between spectral frequencies. The difference is caused by vocal tract length of a training speaker and an input speaker. Recognition of subsequent utterances using this determination improves recognition performance of the subsequent speech sounds.

Growing community of inventors

Kanagawa, Japan

Keiko Morii