Application of neural network for digital recursive filter design

Digital Signal Processing is an advanced technology that will determine the direction of science and technology in the next centuries. One of the main direction of digital signals processing is digital filters, which in the most cases have advantages over analog filters. Currently there are various methods of filter analysis and design. In this work, for synthesis of all types of recursive filters (low pass, high pass, bandwidth, band stop) is used a neural network. The main objective of filter synthesis is to find the filter coefficients. These filter coefficients define the filter transfer function. Using an iterative procedure of the neural network such as Backpropagation algorithm, on base of Visual C++ software was developed the program, which designs recursive filters with required characteristics. This is particularly important for the designing of the new correcting filters characteristics, the purpose of which is to reduce the unwanted noises in the measurement signal.