A Probabilistic Speech Recognition Framework Based on the Temporal Dynamics of Distinctive Feature Landmark Detectors

Aren Jansen; Partha Niyogi. 1 June, 2007.
Communicated by Partha Niyogi.


In this paper, we elaborate on a computational model for speech recognition that is inspired by several different interrelated strands of research in phonology, acoustic phonetics, speech perception, and neuroscience. Our goals are twofold: (i) to explore frameworks for recognition that may provide a viable alternative to the current hidden Markov model (HMM) based speech recognition systems (ii) to provide a computational platform that will allow us to engage, quantify, and test various theories in the scientific traditions in phonetics, psychology, and neuroscience. Our approach uses the notion of distinctive features, constructs a hierarchically structured point process representation based on feature detectors, and probabilistically integrates the firing patterns of these detectors to decode a phonetic sequence. We find the accuracy of a broad class recognizer based on this framework to be competitive with equivalent HMM-based systems. We conclude by outlining various avenues for future development of our methodology.

Original Document

The original document is available in PDF (uploaded 1 June, 2007 by Partha Niyogi).