Research Groups

Speech

The Speech Group (formerly the Realization Group) conducts research in the areas of algorithms, architectures, and systems for speech and audio signal processing and pattern recognition. In the 1990s the emphasis of this group's work gradually became the application of these areas of study to problems in spoken language processing, and particularly in robust automatic speech recognition.

The Speech Group works closely both with other groups within ICSI and with several faculty members from the Electrical Engineering and Computer Science Department within UC Berkeley.

Key research areas of the Speech Group include auditory-inspired signal processing, statistical modeling, rich transcription of natural and impromptu meetings, and rich transcription of broadcast and conversational telephone speech.

Professor Nelson Morgan of the Electrical Engineering Faculty at UC Berkeley leads the Speech Group. With Hervé Bourlard, he was an originator of the "hybrid system" approach to speech recognition (neural networks used probabilistically with HMMs), and with Hynek Hermansky was the co-inventor of signal processing techniques now used in millions of cell phones. He is a Fellow of the IEEE.

Browse Publications of the Speech Group

Read about specific projects of the Speech Group.
Speech research software downloads
Speech Group FTP site
Documents for local users only

top