Speech database and acoustic-phonetic knowledge base for speech recognition

This paper describes a speech database and phonetic knowledge acquiring system developed for speech analysis and recognition research at ETL. Its system design concept, system configuration and some developed techniques are presented as well as acoustic-phonetic knowledges obtained from the speech database. The system is characterized by 1) fine phonetic labeling system and 2) network representation for acoustic-phonetic knowledges. An acoustically compact segment called APSeg is adopted for the labeling of speech samples. The acoustic-phonetic knowledges are automatically derived by sorting the labeling results of a real speech database.