Design of linear-phase FIR filters using pseudo-Boolean methods

To implement a given digital filter with a finite-bit digital hardware, its design parameters have to be rounded off or truncated. The resulting filter characteristics considerably deviate from the desired ones. This effect can be reduced if the parameters are suitably chosen in the discrete-parameter-space. In this paper, this is achieved for linear-phase FIR filters by solving a system of linear pseudo-Boolean inequalities. The method is simple to use and is believed to offer a new insight into the problem. It takes much less computer time as compared to the existing search techniques, and yet has much better chances of converging to a global optimum.