We propose a network architecture which includes an improved pre-processing system cell arrangement, recognition algorithm and learning rule. A speech recognition system using the SC Parallel Distributed Processing network has been proposed. Network construction and tuning procedure have been reported. The tracking method of the local peak pattern has been described. Post-processing system to recognize the input speech without use of any external matching system has been proposed. Recognition results using computer simulation are presented.
|出版ステータス||Published - 1989|
ASJC Scopus subject areas
- Electrical and Electronic Engineering