Journal of the Faculty of Engineering and Architecture of Gazi University 35:1 (2020) 193-211 Design, analysis, and ECG subband processing application of new M-channel cosine modulated uniform filter banks based on exponential window family Kemal Avcı 1* , Eda Gümüşsoy 2 1 Department of Electrical and Electronics Engineering, Izmir Democracy University, Izmir, 35140, Turkey 2 Arcelik A.S., Bolu, 14020, Turkey Highlights: Graphical/Tabular Abstract  Design of new M- channel cosine modulated uniform filter banks with near perfect reconstruction  Detailed and comparative performance analysis of the proposed design method  Comparative performance analysis of the proposed design method on ECG subband processing application Figure A. Block diagram for the proposed method Purpose: Amplitude and aliasing errors inherently occur in designing of M-channel cosine modulated filter banks. In this study, it is aimed to present a new design method for M-channel cosine modulated uniform filter banks to minimize these errors. Theory and Methods: To design M-channel cosine modulated uniform filter banks, exponential window family having two adjustable parameters is used to obtain the prototype filter. To find the optimum cut-off frequency of the prototype filter, the optimization method proposed by Lin and Vaidyanathan to minimize the cost function is used. The performance of the proposed method is measured in terms of amplitude (Ett) and aliasing (Eto,m) errors. To evaluate the performance of the designed method in ECG subband processing application, three parameters, namely PRD (Percent Root-mean-square Difference), MSE (Mean Square Error), and ME (Maximum Error) are used. Comparisons are performed with Kaiser window based Lin’s method. Results: The filter banks designed by the proposed method are first analyzed in terms of amplitude and aliasing errors for various channel number (M), minimum stopband attenuation (As), and filter length (N). The results showed that to minimize the amplitude error, the filter length must be chosen according to the channel number and minimum stopband attenuation. For example, the minimum E tt values for As = 100 dB are obtained for the filter lengths 31 ≤ N ≤ 47 for M = 2, 53 ≤ N ≤ 71 for M = 8, 185 ≤ N ≤ 301 for M = 16, and 385 ≤ N ≤ 601 for M = 32. Also, it is observed that minimum aliasing errors are obtained when the minimum stopband attenuation becomes higher for a fixed filter length. Then, a comparative filter bank design example for M = 32, As = 100 dB, and N = 467 is carried out and results showed the filter bank designed by the proposed method presents less amplitude and aliasing errors than the one designed by the Kaiser window based Lin’s method. As for ECG subband processing application, the proposed method generally provided better results than Lin’s method in terms of PRD, MSE, and ME. Conclusion: A new method based on exponential window family is proposed to design M-channel cosine modulated filter banks. After extensive analysis, the filter banks designed by the proposed method are compared with the ones designed by Lin’s method. It is concluded that the proposed method generally provides better results in terms of E tt and E to,m , and also in terms of PRD, MSE, and ME for ECG subband processing application. Keywords:  Cosine modulated filter bank  Near perfect reconstruction  ECG subband processing  Exponential window  Kaiser window Article Info: Research Article Received: 07.08.2018 Accepted: 16.08.2019 DOI: 10.17341/gazimmfd.451652 Correspondence: Author: Kemal Avcı e-mail: kemal.avci@idu.edu.tr phone: +90 232 2601001