News & Events
Topic: Cancellation of Research Seminar : " Intelligent Speech Processing "
Posted - 22/10/2007 : 12:00:40
City University of Hong Kong Dep
Department of Chinese, Translation and Linguistics
Intelligent Speech Processing
Dr. David M. Horowitz
VoxIQ Limited U.K.
Date: 30 October 2007, Tuesday (Cancelled)
Time: 4:30pm - 6:00pm
Venue: G7619 (Lift 3, 7/F, Green Zone),Academic Building,CityU
This talk will cover two topics: recent work in spoken dialogue understanding and the application of source-filter bio-inspired models for text-to-speech synthesis.
In dialogue understanding, we describe the application of an artificial cognitive system based on knowledge based systems techniques. We discuss how the application of KBS techniques can increase the effective vocabulary size of high performance key word spotting by maintaining exquisite understanding of the context and topic of conversation.
The second part of the talk covers the theory and principles of work performed at MIT which embodied a bio-inspired model of speaker voice quality and prosody. The theory is discussed to motivate the application of certain acoustic characteristics that have been shown to be associated with voice quality. A possible approach is discussed of how these acoustic measures might aid in improving the modelling of speaker characteristics for statistically or concatenative based text-to-speech synthesis.
Dr. Horowitz is a speech technologist with background in both academia and industry. He was principal developer of one of the first multi-modal speech based robotic systems between 1988-1992 at New England Medical Center and Tufts University School of Medicine where he was Associate Director of Rehabilitation Engineering and Research Assistant Professor of Rehabilitation Medicine. For the last decade, Dr. Horowitz has worked at the bleeding edge of research and commercial product. He has held positions at L&H, NetByTel, Vox Generation and currently VoxIQ, where he is Head of Speech Technology. He has been responsible for two industry firsts: VoiceRAD (a radiology speech dictation system) and the L&H Natural Language Interface now being utilized in Dragon Naturally Speaking. Dr. Horowitz is the principal inventor of 8 awarded or pending patents and the author of 25 publications in the area of speech technology. He has been the Principal Investigator of several government grants and contracts. He was awarded a Ph.D. by Publication from the University of East Anglia-Norwich, England and was a graduate student at MIT in the Department of Electrical Engineering and Computer Science where he earned his Sc.M. degree. He performed doctoral research in fundamental speech acoustics applied to elucidating measures of voice quality and prosody.