Speaker: Dr. Mark Clements

Speaker's Title and Affiliation: Professor, Georgia Tech

Seminar Title:
Issues and Solutions in Audio Retrieval

Abstract: The problem of retrieving voice and audio from non-transcribed sources is in its infancy compared to text and meta-data searching. Despite recent advances in speech-to-text (TTS) systems, there are many applications where such approaches are not viable. A phonetic-based high-speed keyword spotting technique was developed at GT as an alternative. It enjoyed sufficiently good success that a commercial enterprise (Nexidia) was established to further develop capabilities and applications. Demos will be given and the technology will be described along with interesting new research results and capabilities.

Speaker Bio:
Dr. Mark A. Clements is professor of electrical and computer engineering at the Georgia Institute of Technology, where he holds the Joseph M. Pettit Endowed Professorship in Digital Signal Processing. His also served as the director of Georgia Tech's Interactive Media Technology Center (IMTC) from 1999-2012. He received the S.B. (Bachelor's), S.M. (Master's), E.E. (Professional Engineer's), and Sc.D. (Doctorate) degrees in 1976, 1978, 1979, and 1982, all in electrical engineering and computer science from Massachusetts Institute of Technology. He is a Fellow of the Institute of Electrical and Electronics Engineers (IEEE), has been a member of the IEEE Speech Technical Committee, has served as an editor for IEEE's Transactions on Acoustics, Speech, and Signal Processing, and was elected to the Signal Processing Society's Board of Governors. Professor Clements is also founder and director of Nexidia, an Atlanta-based speech technology company. Professor Clements' current research interests involve digital processing of speech signals. The research is concerned with such problems as the application of digital speech technology to sensory aids for the hearing impaired and automatic recognition of speech in adverse conditions. Some of the interesting problems arising from these applications include enhancement of speech in noise, formulation of robust perceptual distance measures, and real-time implementation. Dr. Clements also does work in efficient coding of speech signals, auditory modeling for improved speech analysis, speech production modeling, general digital signal processing, and pattern recognition.