基于神经网络的Bark域图形均衡器的设计OACSTPCD

A method based on back propagation(BP)neural network is described to simplify the design of graphic equalizer without sacrifi-cing approximation accuracy.Its core idea is to train a neural network to predict the mapping relationship between the target gain and the optimized bandpass gain at the specified center frequency.In the case of a 24-channel Bark band graphic equalizer,the data fitting func-tion of the BP neural network with a hidden layer of 48 neurons is used to realize the prediction.Then,the closed formula is used to calcu-late the coefficients of the band filter quickly and easily.The precise control method of using a least square method to obtain the optimal gain of the infinite impulse response(IIR)filter is introduced and continued to be improved.BP neural network and target gain are used to obtain the optimal gain of the parameter equalizer,greatly reducing the amount of calculation and making the approximation error less than 0.1 dB.The resulting neural controlled 24-channel Bark domain graphic equalizer is very useful in audio conference equalization requiring time-varying equalization.