18 Mar 2003
Articulatory-acoustic features in ASR - beyond the "beads-on-a-string" paradigm
In current state-of-the-art automatic speech recognition systems, the acoustic signal is usually described in terms of phones, and words are simply seen as concatenations of phone sequences. However, as many before me have pointed out, the notion that a word is composed of a sequence of phone segments, i.e., the "beads-on-a-string" paradigm, is questionable. In this talk, I will present my views on how articulatory-acoustic features in combination with syllables can be used to circumvent the beads-on-a-string issue and at the same time can be used to improve automatic speech recognition.