2-D IIR filter design using an adaptation of the random search methodology

A number of techniques are available for generating low-pass filters [1] that help to remove additive noise from transmitted signals. Frequency transformation techniques [2,3,4] can be subsequently used to convert these filters into those with different specifications. However, they will always result in either high-pass, band-pass or band-reject filter types. However, the need may arise for designing a nonstandard filter of arbitrary specifications. This includes cases where signals in multiple bands may need to be transmitted or rejected at different attenuation levels. In this paper, we discuss a method for designing such a filter with arbitrary frequency response characteristics. The design technique is based on an adaptation of the random search method [5]. The only inputs required for this algorithm are the desired frequency response and resolution. The algorithm will create a frequency domain transfer function of the desired characteristics (both magnitude and phase). A uniform random number generator will be required for the functioning of the algorithm.