Hybrid Filtering : Frequency Domain Approximate Filtering and Time Domain Compensation
Yoshiki Kutsuwa, Ukeji Nobutaka, Koichi Ichige, Rokuya Ishii · Institutional Repositories DataBase (IRDB) · 2009
Abstract—This paper presents an efficient FIR filtering method using frequency domain approximate processing and time do-main compensation. Most of basic operations in signal/image processing are based on FIR filtering, i.e., discrete convolution. Fast Fourier Transform (FFT) can be used to implement discrete convolution in frequency domain in order to reduce compu-tational complexity, for certain signal and impulse response durations. However FFT often needs zeros to be added to signals, since it is designed for the signals whose duration is power of two. Therefore the operation sometimes contains many redundant calculations. This paper presents a computationally efficient filtering method using modified frequency domain processing and compensation in time domain which can improve efficiency in convolution operation. Its effectiveness is verified through some computer simulation and DSP implementation. I.