ANALOGUE ELECTRONIC CIRCUIT FOR ARTIFICIAL NEURON MODELLING

摘要

The present invention relates to a speech recognition using an analogous neural network, sensitive to the concurrency degree of events, which may be trained both for detecting words by association of successive sounds and for detecting sounds by the association of formants specific thereto, the neural network being based on a hardware-implemented biological-like neuron model, using only micro-technologically integrable components, the neural network maintaining the properties of natural neurons and being trained for the recognition of sounds and words by a mechanism of stimulating pre- and post-synaptic neuron by using supraliminal stimuli.