Study design manner of the waveform processing device by the neural network

摘要

PURPOSE:To provide a method for designing a waveform equalizer capable of accurately reproducing zero crossing by means of a neural network (NN). CONSTITUTION:In the designing method for the learning of a waveform equalizer by means of the NN 11, the NN 11 is learned so that the weights of plural conversion systems are changed and converged by reducing an error (f) between a required signal value (d) and an output value (c) of the NN 11 in learning and the error (f) is largely evaluated (e is expressed by E) in the vicinity of the zero crossing of the value (d). The vicinity of the zero crossing of the value (d) is a prescribed section M before and after the time of the zero crossing point (m) of the value (d) or an interval (N) between the zero crossing point of the value (d) and the zero crossing point (n) of the output value from the NN 11 in learning.