System and chip design for neural and auditory signal processing

Speech recognition and understanding is one of the most difficult tasks to be carried out by machines while it is readily performed by humans. Although modern electronic and computer engineering has greatly advanced in recent years, construction of a perfect electronic speech recognition system, i.e. inexpensive, fast, speaker-independent, capable of handling continuous speech, and exhibiting high recognition rate still eludes our endeavour. In this communication we describe an effort to build an advanced speech recognition system that combines the most recent progress in psychoacoustics, auditory perception, neural networks, and silicon technologies.