The goal of this research is development of a new database to aid researchers in automatic recognition of continuous speech. Dr. Cole will conduct controlled experiments to determine the effect of phonetic context on human recognition of speech excerpts. Initial studies will make use of phonetically labeled excerpts from the widely used TIMIT speech database. Human phonetic recognition will be measured with varying retention of preceding and following phonetic cues. Dr. Cole will also attempt to design neural-network classification algorithms that match human phonetic classification performance.
