Biophysical recognition system of time-series analog signals by a hebbian learning network

We propose a biophysical recognition system of time-series analog signals based on fundamental neural information processing. The system utilizes the following process: (i) the time-series analog signals are altered to an impulse train by multi-kinds or multi-channels of transducers in a transducing process, (ii) the produced impulse train propagates within a biophysical neural network in keeping with the frequency of the impulse train, (iii) the time-series of final output patterns through the biophysical neural network are compared with samples stored in memory, and recognition and discrimination are performed according to the maximum similarity between patterns. The ability of the proposed system is verified by using artificial analog signals and utterance signals.