1, (1) draw the speech signal speech1_10k. Wav of time domain waveform, the signals from a approximate periodic vowels, duration of 25 ms, sampling rate of 10000 hz,
(2) with 1024 point FFT, draw the speech signal in the range 0,?????? The Fourier transform of the logarithm amplitude curves, request signal with weighted hamming window, window respectively take 25 ms and 10 ms, hamming window is located in the center of the signal, draw two window length of logarithmic amplitude curves,
(3) according to the speech signal waveform, it is estimated that the cycle of quasi-periodic signal in seconds (pitch), which shows that 25 ms and 10 ms window spectrum estimation is more suitable for pitch estimation? Why is that? Try to use your knowledge of the pitch estimation based on the cepstrum method and weighted hamming window, window respectively take 25 ms and 10 ms, hamming window is located in the center of the signal, draw two window length of logarithmic curve,

CodePudding user response:

Really beg the heart tired