Search

Pronunciation by analogy (PbA) is a data-driven approach to phonetic transcription that generates pronunciations for unknown words by exploiting the phonological knowledge implicit in the dictionary that provides the primary source of pronunciations. Unknown words typically include low-frequency ‘common' words, proper names or neologisms that have not yet been listed in the lexicon. It is received wisdom in the field that knowledge of the class of a word (common versus proper name) is necessary for correct transcription, but in a practical text-to-speech system, we do not know the class of the unknown word a priori . So if we have a dictionary of common words and another of proper names, we do not know which one to use for analogy unless we attempt to infer the class of unknown words. Such inference is likely to be error prone. Hence it is of interest to know the cost of such errors (if we are using separate dictionaries) and/or the cost of simply using a single, undivided dictionary, effectively ignoring the problem. Here, we investigate the effect of lexicon composition: common words only, proper names only or a mixture. Results suggest that high-transcription accuracy may be achievable without prior classification. [ABSTRACT FROM AUTHOR]

In previous work, we have confirmed the performance gains that can be obtained in speaker recognition by splitting the (clean) wide-band speech signal into several subbands, employing separate pattern classifiers for each subband, and then using multiple classifier fusion (`recombination΄) techniques to produce a final decision. However, our earlier work used fairly rudimentary recognition techniques (dynamic time warping), just sum or product fusion rules and the spoken word seven only. The question then arises: Can subband processing still deliver performance gains when using state-of-the-art recognition techniques, more sophisticated recombination, and different spoken digits? To answer this, we have applied hidden Markov modelling and artificial neural network (ANN) recombination to text-dependent speaker identification, for spoken digits seven and nine. We find that ANN recombination performs about as well as the sum rule operating in log probability space, but the ANN results are not unique. They depend critically on user-specified parameters, initialisation, etc. On clean speech, all classifiers achieve close to 100% identification. Subband techniques offer advantages when the speech signal is significantly degraded by noise. [ABSTRACT FROM AUTHOR]