![]() |
Workshop on the Auditory Basis of Speech PerceptionKeele University, UK |
![]() |
The first decade of work into computational auditory scene analysis (CASA) of speech has been dominated by approaches which regard auditory organisation as a preprocessor for speech recognition. In this paper, I will argue that for CASA to progress beyond speech enhancement is no simple task and will require the development of a computational theory of speech perception which may look quite different from the accounts provided by the best performing recognition scheme we possess at present, based on hidden Markov models. Some of the criteria which an adequate, perceptually-inspired model needs to account for are outlined, and these are contrasted with those which currently motivate robust automatic speech recognition systems. Modelling studies addressing some of the competencies required for a computational approach to speech perception are described. One study considers the requirement to cope with missing data. Another concerns the role of counter-evidence suggested by principles of perceived auditory continuity. The paper concludes by considering explicit aspects of the integration of auditory organisation and speech perception.
Bibliographic reference. Cooke, Martin (1996): "Auditory organisation and speech perception: arguments for an integrated computational theory", In ABSP-1996, 186-193.