Signal Processing and Recognition in Adaptive Neural Networks.

摘要

The research reported here has been conducted over the last three years under the AFOSR grant (AFOSR-88-0204). It was divided into four general categories of projects: (1) Cochlear models: applications and implementations. Detailed models of the auditory periphery were developed and applied as front-ends for speech recognition experiments. (2) Early auditory processing: binaural hearing and phonemic segmentation. Physiological plausibility of traditional models was examined and alternative formulations were made and tested. (3) Central auditory function: physiology, psychoacoustics, and mathematical models. Experiments focused on the primary auditory cortex and the anterior auditory field. Models of the responses were applied to generalized representations of speech. Psychoacoustical experiments were carried out to elaborate and test the physiologically derived models. (4) Analysis of neural networks in applications to tactile sensing. Mathematical formulations of the deconvolution problem were analyzed and solved using neural network structures. (Author)