Linear Phase FIR Filter Design for Digital Hearing Aids Using a Neural Network-Based Optimization

Abstract

A neural network-based optimization for the design of linear phase digital filters used in digital hearing aid applications is presented in this paper. To achieve hearing loss compensation, a 53-tap finite impulse response (FIR) filter is utilized, and the weights of the trained network can be used to obtain the impulse response of the FIR filter. The target frequency response or label comes from the audiogram, the training data is created, and a linear perceptron supervised learning model is used. Audiogram matchings are shown in design examples. The proposed neural network-based design approach can give a very high accuracy, high processing speed when compared with the existing methods, and also low complexity in filter structure realization.