Matsushita Electric Industrial Co., Ltd.

Katsuyuki Niyada

Taisuke Watanabe

Shoji Hiraoka

Satoru Fujii

Masakatsu Hoshimi

Shuji Morii

Maki Yamada

Ikuo Inoue

Tatsuya Kimura

Katsuhiko Noda

Kenji Kaga

Maki Miyata

An inter-frame similarity between an input voice and a standard patterned word is calculated for each of frames and for each of standard patterned words, and a posterior probability similarity is produced by subtracting a constant value from each of the inter-frame similarities. The constant value is determined by analyzing voice data obtained from specified persons to set the posterior probability similarities to positive values when a word existing in the input voice matches with the standard patterned word and to set the posterior probability similarities to negative values when a word existing in the input voice does not match with the standard patterned word. Thereafter, an accumulated similarity having an accumulated value obtained by accumulating values of the posterior probability similarities according to a continuous dynamic programming matching operation for the frames of the input voice is calculated for each of the standard patterned words. Thereafter, a particular standard patterned word relating to an accumulated similarity having a maximum value among the accumulated similarities is output as a recognized word of the input voice.

Voice recognition method for recognizing a word in speech

A set of 'm' feature parameters is generated every frame from reference speech which is spoken by at least one speaker and which represents recognition-object words, where 'm' denotes a preset integer. A set of 'n' types of standard patterns is previously generated on the basis of speech data of a plurality of speakers, where 'n' denotes a preset integer. Matching between the feature parameters of the reference speech and each of the standard patterns is executed to generate a vector of 'n' reference similarities between the feature parameters of the reference speech and each of the standard patterns every frame. The reference similarity vectors of respective frames are arranged into temporal sequences corresponding to the recognition-object words respectively. The reference similarity vector sequences are previously registered as dictionary similarity vector sequences. Input speech to be recognized is analyzed to generate 'm' feature parameters from the input speech. Matching between the feature parameters of the input speech and the standard patterns is executed to generate a vector of 'n' input-speech similarities between the feature parameters of the input speech and the standard patterns every frame. The input-speech similarity vectors of respective frames are arranged into a temporal sequence. The input-speech similarity vector sequence is collated with the dictionary similarity vector sequences to recognize the input speech.

Method of speech recognition

A method of speech recognition includes the steps of analyzing input speech every frame and deriving feature parameters from the input speech, generating an input vector from the feature parameters of a plurality of frames, and periodically calculating partial distances between the input vector and partial standard patterns while shifting the frame one by one. Standard patterns correspond to recognition-object words respectively, and each of the standard patterns is composed of the partial standard patterns which represent parts of the corresponding recognition-object word respectively. The partial distances are accumulated into distances between the input speech and the standard patterns. The distances correspond to the recognition-object words respectively. The distances are compared with each other, and a minimum distance of the distances is selected when the input speech ends. One of the recognition-object words which corresponds to the minimum distance is decided to be a recognition result.

In a voice recognition method, a d-by-J demensioned reference voice pattern is prepared for each target word, when J denotes a predetermined number of frames and d denotes a predetermined number of characterizing parameters per frame. A spoken input word is partitioned between its start and end points into J frames, and d characteristic parameters are extracted for each frame to form a d-by-J demensioned input time-series vector. The resemblance between the input vector and each of the reference voice patterns is then calculated using a statistical distance scale, and the spoken word is identified with the reference pattern providing the highest resemblance. The method requires fewer calculations and yet attains a high recognition rate through the normalization of the input voice word for both spectrum and time.

Voice recognition method

Linear prediction coefficients of a speech signal including unknown words are derived for each of successive periodic frame intervals. For every frame over the duration of an individual phoneme of the speech signal, the degree of similarity of stored coefficients of known words and derived coefficients of the unknown words are calculated so that at the end of the individual phonemes, the degree of similarity is calculated. Phoneme segmentation data are derived in response to the speech signal and combined with the calculated degree of similarity over the individual phoneme to derive phoneme strings of the speech signal. The derived and stored phoneme strings are compared to indicate the words stored in a word dictionary having the greatest similarity with the derived phoneme strings.

Method of and apparatus for speech recognition wherein decisions are

Phoneme standard patterns are prepared beforehand using speech sounds from a number of individual speakers. Unknown input speech sounds are divided into continuous frames and then some of these frames are extracted such that extracted frames include noncontinuous frames. LPC cepstrum coefficients are obtained for each of the frames and a mean value is obtained from the LPC cepstrum coefficients for each of the phonemes. A first standard pattern is then formed for each of the phonemes belonging to each group determined by characteristics of known speech sounds. Spectrum information of the unknown input speech sounds is produced using such extracted frames, and is then compared with the phoneme standard patterns prestored in a storage to determine and recognize phonemes of the unknown speech sounds by calculating similarity between the two using a statistical distance measure. When obtaining LPC cepstrum coefficients through LPC analysis, the order of LPC cepstrum coefficients is set to a value below the order of LPC analysis. The method according to the present invention reduces the amount of calculations necessary to perform speech recognition without deteriorating phoneme and word recognition rate.

Method for speech recognition

Disclosed is a speech recognition apparatus comprising: a speech analysis portion for extracting parameters necessary for determination of spoken words; a speech period detecting portion for extracting one or more combinations of speech periods using the parameters; and a structure analysis portion for detecting feature points indicative of phoneme structure of each word and for determining a word through computation of similarity to proposed words in accordance with the presence and absence of the feature points. Therefore, erroneous recognition due to noise introduction or the like can be reduced by detecting one or more combinations of proposed speech periods by the speech period detecting portion. By extracting only necessary number of extracting points, which contribute to the distinguishment between words, with reference to analysis procedure provided for each word, the sharpness of determination is bettered. More stable operation than conventional apparatus has been achieved in connection with time base expansion/compression. Small numbers of parameters obtained through speech analysis are used to reduce the amount of computation, while the above-mentioned parameters are stable against difference in phenemes due to difference in speakers.

Apparatus for speech recognition

Speech parameters (P.sub.h and P.sub.l) are derived for consonant classification and recognition by separating a speech signal into Low and High frequency bands, then in each band obtaining the time first-derivative, from which the min-max differences (power dip) are obtained (P.sub.h and P.sub.l). The distribution of P.sub.h and P.sub.l in a two-dimensional plot for a discriminant diagram classifies the consonant phoneme.

Method and apparatus for speech recognition

Apparatus for speech recognition, having each phoneme as a fundamental recognition unit, recognizes input speech by discriminating phonemes in the input speech. The apparatus comprises a memory for storing phoneme standard patterns of phonemes or phoneme groups; a spectrum analyzer for obtaining parameters indicative of the input speech signal spectrum; a statistical distance measure similarity calculator calculates the degree of similarity between the output of the spectrum analyzer and standard patterns stored in the memory; a segmentation portion for segmenting by using time-dependent low- and high-frequency power variations of the input speech signal and results from the similarity calculator; and a phoneme discriminator for recognizing phonemes by using the results from the similarity calculator.

In this speech synthesizer which announces the time, storage and complexity is minimized by switching selection of fixed data and variable data from corresponding memories, to compose the phrase 'the time is five o'clock'.

Sound synthesizing apparatus

A single section digital filter is provided with a digital input signal and includes a serial multiplier having an input responsive to predetermine parameters. An adder/subtractor has an input serially connected to the multiplier output. A first switch serially selects recirculating data from an adder/subtractor output directly or delayed, or the digital input signal. A second input of the multiplier is serially provided with the selected data. A second switch serially and selectively connects delayed selected data to a second input of the adder/subtractor. A latch serially receives delayed data for effecting series/parallel conversion thereof.

Digital filter for performing serial operations and vocal sound

An input signal is applied to a clamping circuit where the level of the input signal is clamped at a desired base line voltage in response to a clamping control signal. The clamping control signal is a pulse train signal, and pulses are applied to the clamping circuit only when the level of the output signal of the clamping circuit is within a given range. Therefore, when the level rising rate or speed is low, the level of the input signal is intermittently clamped so that the output signal level is maintained close to the base line voltage. When the level rising rate exceeds a given value, clamping is not performed so that the output level follows the input level. In order to see whether the level of the output signal is within the given range or not, a reference voltage which is higher than the base line voltage is used when shaping a two-level signal. If the input signal is of three-level, another reference voltage, which is lower than the base line voltage, is additionally used.

Method and circuit arrangement for shaping a signal waveform

An electronic engraving and recording system comprising a television camera for picking up an image of an object and converting this image into an electrical signal, means for generating control signals on the basis of the synchronizing signal used in the television camera, memory means for storing the electrical signal under control of the control signals, means for engraving and recording the image according to the signal read out from the memory means under control of the control signals, and monitoring display means for displaying the visible image of the object in response to the application of the signal read out from the memory means. An image of an object can be simply engraved and recorded on a card within a short period of time without requiring any photographic original of the object.

Growing community of inventors

Sagamihara, Japan

Katsuyuki Niyada