A new FIR filterbank design algorithm using frequency scaling and modulation

Abstract

Digital filterbanks, an important signal processing tool, are used in a radar signal processor to measure target range and/or velocity [4]. The windowed FFT is the most common radar filterbank. The FFT is a special case of modulated filterbank. Modulated filterbanks subdivide the frequency region by modulating a prototype baseband filter. This baseband filter usually satisfies certain frequency response constraints. Filterbanks used in radar applications are not concerned with reconstruction as are wavelet-based approaches [2], but rather with how well the filters subdivide the frequency region into a prescribed number of equal frequency regions. Wavelets use time decimation or frequency scaling to form filterbanks. This paper focuses on using a wavelet-like scaling approach along with modulation to design a digital filterbank. However, we are not concerned with reconstruction but with the subdivision of the frequency region into a prescribed number of equal segments. It will be shown that this new design algorithm has a frequency response performance comparable to existing techniques but with less computational effort. The performance measures considered are rejection of out of bands signals, uniformity of inband signal gain, and transition width from inband to stopband.