Asahi Kasei Kabushiki Kaisha

Yoji Ishikawa

Toshiyuki Miyazaki

Each word to be recognized is represented by gender-specific hidden Markov models that are stored in a ROMalong with output probability functions and preset transition probabilities. A speech recognizerdetermines an occurrence probability of a feature parameter sequence detected by a feature value detectorusing the hidden Markov models. The speech recognizerdetermines the occurrence probability by giving each word a state sequence of one hidden Markov model common to the gender-specific hidden Markov models, multiplying each preset pair of an output probability function value and a transition probability together among the output probability functions and transition probabilities stored in the ROM, selecting the largest product as the probability of each state of the common hidden Markov model, determining the occurrence probability based on the selected product, and recognizing the input speech based on the occurrence probability thus determined.

Speech recognition device and speech recognition method

Signal C&equals;&lsqb;C,, C,, . . . , C,, C,&rsqb; generated by a ring counter,is converted at a logic circuit,to a triangle pulse signal Q in which the carrier frequency of the PWM signal is equal to the frequency of the signal C raised to power of 2. The signal Q is input to the B input of a magnitude comparator,. On the other hand, the modulation data D held by a data holding circuit,is compared with the converted signal Q in the magnitude comparator,, and when the data D is greater than the signal Q, a PWM signal is obtained from the comparison result A>B. This PWM signal has no phase change and high resolution and is a low pulsating component. The resolution of the PWM signal per cycle of the ring counter,is held at n bits, and the carrier frequency is equal to the ring counter frequency raised to mth power of 2.

Growing community of inventors

Machida, Japan

Yoji Ishikawa