Speaker:  Asif Ali

Title:
Voice Query-by-Example System using an Ergodic Hidden Markov Model of Speech

Abstract:
An ergodic hidden Markov model (EHMM) can be useful in extracting underlying structure embedded in connected speech without the need for a time-aligned transcribed corpus.  In this talk, we present a query-by-example (QbE) spoken term detection system based on an ergodic hidden Markov model of speech. Due to the unsupervised nature of the training, a single phoneme may be mapped to a number of EHMM states. The effects of speaker-dependent and context-induced variation in speech on its EHMM-based representation have been studied and used to devise schemes to minimize these variations. A search framework, consisting of a keyword modeling scheme and a modified Viterbi algorithm, has also been implemented. The evaluation metrics for the EHMM-based QbE system on the MediaEval2013 Spoken Web Search task are presented and compared with systems of similar complexity based on static clustering schemes.

Bio:
Asif Ali received B.S. and M.S. degrees in electrical engineering from the University of Engineering and Technology, Lahore, in 2001 and 2003, and M.S. degree in signal processing and communication from the University of Edinburgh in 2005. He is currently working toward his Ph.D. in electrical and computer engineering, under the supervision of Professor Mark A. Clements, at the Georgia Institute of Technology. His area of research is speech recognition systems for resource-limited languages.