A motor theory of speech perception, initially proposed to account for results of early experiments with synthetic speech, is now extensively revised to accommodate recent findings, and to relate the assumptions of the theory to those that might be made about other perceptual modes. According to the revised theory, phonetic information is perceived in a biologically distinct system, a ‘module ’ specialized to detect the intended gestures of the speaker that are the basis for phonetic categories. Built into the structure of this module is the unique but lawful relationship between the gestures and the acoustic patterns in which they are variously overlapped. In consequence, the module causes perception of phonetic structure without translation from preliminary auditory impressions. Thus, it is comparable to such other modules as the one that enables an animal to localize sound. Peculiar to the phonetic module are the relation between perception and production it incorporates and the fact that it must compete with other modules for the same stimulus variations.

Abstract not found

A typical auditory scene in a natural environment contains multiple sources. Auditory scene analysis (ASA) is the process in which the auditory system segregates a scene into streams corresponding to different sources. Segmentation is a major stage of ASA by which an auditory scene is decomposed into segments, each containing signal mainly from one source. We propose a system for auditory segmentation by analyzing onsets and offsets of auditory events. The proposed system first detects onsets and offsets, and then generates segments by matching corresponding onset and offset fronts. This is achieved through a multiscale approach. A quantitative measure is suggested for segmentation evaluation. Systematic evaluation shows that most of target speech, including unvoiced speech, is correctly segmented, and target speech and interference are well separated into different segments.

Computational auditory scene analysis -- modeling the human ability to organize sound mixtures according to their sources -- has experienced a rapid evolution as the simple principles suggested by psychological experiments have turned out to be less than the whole story. Phenomena such as the continuity illusion and phonemic restoration show that the brain is able to use a wide range of knowledge-based contextual constraints when interpreting obscured or complex mixtures: To model such processing, we need architectures that operate by confirming hypotheses about the observations rather than relying on directly-extracted descriptions. One such architecture, the `prediction-driven' approach, is presented along with results from its initial implementation. This architecture can be extended to take advantage of the high-level knowledge implicit in today's speech recognizers by modifying a recognizer to act as one of the `component models' which provide the explanations of the signal mixtur...

phone:617-353-7857 fax:617-353-7755

This paper contains very many

a neural network model of auditory scene analysis and source segregation