Title :
Combination of two methods for isolated word speech recognition
Author :
Savage, Jesus ; Holden, A.
Author_Institution :
Human Interface Technology Lab., Washington Univ., Seattle, WA, USA
Abstract :
The authors present a method for isolated word speech recognition combining hidden Markov models (HMM) and multisection vector quantization (MSVQ) code books techniques. The use of these two techniques together enables the system to correct half of the wrong decisions made when only one technique is used. This system was implemented on a digital signal microprocessor (DSP), and is operating in real time. This system will be used in a virtual reality environment
Keywords :
Markov processes; digital signal processing chips; speech coding; speech recognition; vector quantisation; virtual reality; DSP; digital signal microprocessor; hidden Markov models; isolated word speech recognition; multisection vector quantization code books; virtual reality environment; Books; Distortion measurement; Filters; Hidden Markov models; Humans; Isolation technology; Linear predictive coding; Shape measurement; Speech recognition; Vector quantization;
Conference_Titel :
Robot and Human Communication, 1992. Proceedings., IEEE International Workshop on
Conference_Location :
Tokyo
Print_ISBN :
0-7803-0753-4
DOI :
10.1109/ROMAN.1992.253928
