Lecture6 FFT.pdfVIP

i i l Si l iD g ta gna Process ng SEE of SWJTU G.X.Zhang zhgxdylan@126.com Chap.7 FFT ? Efficient computation of the DFT: fast Fourier transform (FFT) algorithms ? Direct computation of the DFT ? Divide-and-conquer (分而治之) approach to computation of the DFT ? Radix-2 FFT algorithms (基数) R di 4 FFT l ith? a x- a gor ms ? Split-radix FFT algorithms (分裂基) ? Applications of FFT algorithms SEE of SWJTU2 Spring, 2010 G.X. Zhang Chap.7 FFT ? Importance of DFT ? Many applications in DSP, such as linear filtering, correlation analysis, spectrum, etc. ? The reason: the existence of FFT ? Focusing on the divide-and-conquer approach in which a DFT of size N N it b (可分解的 合数)? — a compos e num er , ? Discussing FFT for computing the DFT when the size N is a power of 2 and 4 SEE of SWJTU3 Spring, 2010 G.X. Zhang 7.1 FFT algorithms ? Complexity of DFT 1N? ? DFT 0 ( ) ( ) , 1knN n X k x n W k N ? ? ?????? ? ? ?? ? IDFT 11( ) ( ) , 1 N kn Nx n X k W n NN ? ?? ?????? ? ? ?? 0k? 2 /j NW e ???where x(n)—complex valued N ? Question 1— Is it necessary to analyze DFT and IDFT individually? No. The same type of computation. SEE of SWJTU4 Spring, 2010 G.X. Zhang 7.1 FFT algorithms ? According to the symmetric properties of DFT 1 2 2N k k? ? ? 0 1 ( ) ( ) cos ( )sinR R I n N n nX k x n x n N N k k ? ? ? ? ? ?? ?? ? ? ? ? 0 2 2( ) ( )sin ( )cosI R I n n nX k x n x n N N ? ?? ? ? ? ? ?? ?? ? ? ? According to the symmetric properties of IDFT 11 2 2N kn kn? ?? ? ? 0 1 ( ) ( ) cos ( )sin 1 2 2 R R I k N x n X k X k N N N k k ? ? ?? ?? ? ? ? ? 0 ( ) ( )sin ( ) cosI R I k n nx n X k X k N N N ? ?? ? ? ?? ?? ? ? SEE of SWJTU5 Spring, 2010 G.X. Zhang 7.1 FFT algorithms ? Question 2— How many complex multiplications and additions do the DFT need? (1) N complex multiplications (4N real multiplications); (2) N-1 complex additions (4N-2 real additions); (3) Totally N values of DFT: N 2 complex multiplications and N 2 -N complex additions ? Therefore, direct c