Abstract
It is important to develop a real‐time continuous speech recognizer for investigating the man‐machine interface through voice. From this point of view, a continuous speech recognizer that employs general digital processors and an augmented continuous DP matching (DTW) algorithm was developed. This equipment consists of an FFT processor for feature parameter extraction, 4 ImPPs (Image Pipelined Processor, NECuPD7281) for DP matching, and a host computer (NEC PC‐9801E) for handling the syntactic constraints. It can be applied to an arbitrary task that is composed of 140 reference patterns and is represented by a finite state automaton with 50 states. Additionally, the method of expression of the automaton for a given task and the semantic extraction to derive the semantic information from the recognition results were designed to improve the flexibility of the user interface. The recognizer with this user interface was applied to the Japanese‐LOGO (vocabulary size: 104 words). In speaker‐dependent recognition tests, the recognizer achieved a sentence accuracy of about 90%.
Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
