In [6] , we develop statistical model of speech recognition where emphasis is placed on the perceptually-relevant and information-rich portion of the speech signal. In that model, speech is viewed as a sequence of elementary decisions or Auditory Events (avents) that are made in response to loci of signicant spectral change. These decision points are interleaved with periods during which insucient information has been accumulated to make the next decision. We h a v e called this a Stochastic

more »

... ceptual Avent Model, or SPAM. In the work reported here, we h a v e extended our initial experimental implementation [7] to include other probabilistic dependencies specied in the original theory, particularly the dependence on the time from the current frame back t o t h e previous hypothesized avent.