VISUAL OBJECT RECOGNITION SYSTEM

摘要

A method and apparatus for automatic object or speech recognition, for example, which uses an algorithm for the unsupervised training of hierarchical feedforward neural networks to perform transform invariant visual object recognition, referred to herein as 'continuous transformation (CT) learning'. This mechanism is based on continuous synaptic enhancement of the feedforward inter-layer connection weights using an associative (such as Hebbian) learning rule during continuous transformation (e.g. translation, rotation, etc) of the (e.g. visual) stimulus.