Abstract:

The design of FIR (Finite Impulse Response) filters is a key issue in digital signal processing. FIR filters make it possible to separate a part of the spectrum that is useful from signals that are undesirable. They are used in virtually every telecommunications device. The problem in designing FIR filters is the time needed to find the desired Filter coefficients so that only the appropriate part of the spectrum is passed with a given accuracy.

The presented paper describes the possibility of generating FIR (Finite Impulse Response) filter coefficients using Deep Neural Networks. This paper will present the neural network structure used for this purpose, an example learning set and the results.