Technical discussions related to Speech Coding (all itu and other vocoders, ACELP, CELP, AMR, etc)
Hi: Which is a better LPC analysis window? Hamming or Hanning? Or, it doesn't matter? Regards, ~Arijit ------------------------------------
| Hi, windowing is not a factor when you do LP analysis. The estimate of spectral envelope obtained using LP analysis is not affected by the discontinuity at the window boundary. So you can just use a rectangular window. The size of window matters somewhat. The spectral envelope remains mostly similar when you choose the window size to be one, two or three pitch cycles (say, 10 ms, 20 ms or 30 ms) for a typical male voice. The position of the window is also not significant here. However, if you go below one pitch cycle, then (a) You have less samples to compute autocorrelation coeffs in LP analysis. (b) It depends on which region of pitch cycle you are choosing to analyze. So, placement of window is important now. The shape of window matters when you do short-time spectral analysis using DFT. Also, the magnitude spectrum obtained from one, two and three pitch cycles will have similar envelopes, but different fine structures. Bye Guru --- On Thu, 5/15/08, #ARIJIT BISWAS# <a...@pmail.ntu.edu.sg> wrote: From: #ARIJIT BISWAS# <a...@pmail.ntu.edu.sg> |