Cepstral Liftering of speech

Spectral smoothing effects of low frequency cepstral liftering on log magnitude spectrum of speech

Speech processing designates a team consisting of Prof. Lawrence Rabiner (Rutgers University and University of California, Santa Barbara), Prof. Ronald Schafer (Stanford University), Kirty Vedula and Siva Yedithi (Rutgers University). This exercise is one of a set of speech processing exercises that are intended to supplement the teaching material in the textbook “Theory and Applications of Digital Speech Processing” by L R Rabiner and R W Schafer.
This MATLAB exercise shows the spectral smoothing effects of low frequency cepstral liftering on the log magnitude spectrum of a speech signal. This is done by first computing the real cepstrum of a window-weighted frame of speech and saving the resulting log magnitude spectrum as the baseline (unsmoothed) speech spectrum. A low frequency lifter is then used to effectively smooth the log magnitude spectrum, with the cutoff frequency varied from a low value of 20 to a high value of 100 in steps of 20 frequencies. The most smoothing occurs with low cutoff frequencies for the lifter; the least smoothing occurs with high cutoff frequencies for the lifter.