Abstract
The authors present windowing techniques and window optimization in the design of resonator-based logarithmic filter banks. The filters in a final filter bank are obtained by weighted sums of elementary filters of a 'mother' bank, and the weights are called windowing weights. The resulting filter bank with improved performance is optimized using linear programming, which is possible since the output response is a linear function of the weights. Examples of window optimizations and bank designs are given, and some trade-offs between the windows and the final filter banks are discussed. From the calculation results, it is found that final filter banks with much better frequency response performance are achieved by optimal windowing.

This publication has 5 references indexed in Scilit: