Progress towards speech models that model speech

Abstract

This paper presents a personal view of recent advances in automatic speech recognition. The analysis is concerned with progress in speech pattern modelling, rather than recogniser performance. Despite the limitations of current approaches, it is argued that extension and development of these techniques provides a viable way forward. It is further suggested that the significance of a number of recent developments, such as sub-band speech recognition and segment modelling, is primarily in their potential for overcoming fundamental limitations of current HMM-based approaches, and not in the short-term improvement in recognition accuracy which has been achieved.