Fast two-level-dynamic-programming algorithm for speech recognition

A three-stage architecture for speech recognition is presented; it includes preprocessing, phoneme recognition, and natural language postprocessor. Within this context of phoneme-based utterance recognition, the paper focuses on the often problematic speed of the second stage and reengineers a standard two-level-dynamic-programming (TLDP) approach to achieve an increase in speed of 75%. Our fast two-level-dynamic-programming algorithm (FTLDP) uses a phoneme clustering technique to reduce the reference search space and silence detection to reduce the length of the utterance to recognize. An overview of the FTLDP algorithm is presented as well as some results.