Neural Networks for Vision, Speech and Natural Language

This book is a collection of papers by British Telecom researchers and their BT funded academic collaborators in the BT Connex project. This project concerns the application of neural networks to image processing, speech technology and natural language processing.

Inhalt

Introduction; Vision; Neural Networks for Vision: an introduction; Image feature location in multiresolution images using a hierarchy of multilayer perceptrons; Training multilayer perceptrons for facial feature location: a case study; The detection of eyes in facial images using radial basis functions; Training and testing of neural net window operators on spatiotemporal images; A neural network feature detector using a multi-resolution pyramid; Image classification using Gabor representation.Speech; Neural Networks for speech processing: an introduction; Spoken alphabet recognition using multilayer perceptrons; Speaker independent vowel recognition; Dissection of perceptron structures in speech and speaker recognition; Segmental subword unit classification using a multilayer perceptron; Natural language; Connectionist natural language processing: an introduction; A single layer higher-order nueral net and its application to context-free grammer recognition; Functional compositionality and soft preference rules; Application of multilayer perceptrons in text-to-speech synthesis systems. Implementation; Hardware imlementation of neural networks: an introduction; Finite word length MLPs; A VLSI architecture for implementing neural networks with on chip back-propagation learning; An opto-electronic neural network processor; Architecture ; Architecture: an introduction; A dynamic topology net; The stochastic search network; Node sequence networks; Some dynamical properties of neural networks. Index.