18.07.2013 Views

LR Rabiner and RW Schafer, June 3

LR Rabiner and RW Schafer, June 3

LR Rabiner and RW Schafer, June 3

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

DRAFT: L. R. <strong>Rabiner</strong> <strong>and</strong> R. W. <strong>Schafer</strong>, <strong>June</strong> 3, 2009<br />

498 BIBLIOGRAPHY<br />

8.17 (MATLAB Exercise) Write a MATLAB program to demonstrate<br />

phase unwrapping for computation of the complex cepstrum of a<br />

speech file. You must carefully compensate any linear phase component<br />

before attempting to do phase unwrapping. Choose a section<br />

of a voiced speech signal <strong>and</strong> compute its DFT <strong>and</strong> decompose it<br />

into the log magnitude <strong>and</strong> phase components (this is the wrapped<br />

phase). Use the MATLAB toolbox comm<strong>and</strong> for phase unwrapping<br />

to unwrap the phase. Compute the complex cepstrum, lifter the<br />

cepstrum using either a low or high quefrency lifter <strong>and</strong> transform<br />

the liftered signal back to the frequency domain <strong>and</strong> ultimately to<br />

the time domain. Plot the following quantities on a single page:<br />

1. the original <strong>and</strong> cepstrally smoothed (in the case of a low quefrency<br />

lifter) log magnitude spectra<br />

2. the original (wrapped) phase, the unwrapped phase, <strong>and</strong> the<br />

cepstrally smoothed phase<br />

3. the complex cepstrum calculated with a large size FFT<br />

4. the resulting estimate of the vocal tract response (obtained by<br />

low quefrency liftering) or the excitation signal (obtained by<br />

high quefrency liftering)<br />

Use the speech file test 16k.wav with starting sample 13,000 <strong>and</strong><br />

frame size 40 msec to test your program. Use an FFT size of 4096<br />

samples <strong>and</strong> a lifter cutoff of 50 samples. Plot the results for both a<br />

low quefrency <strong>and</strong> a high quefrency lifter.<br />

8.18 ( MATLAB Exercies) Write a MATLAB program to illustrate the<br />

effects <strong>and</strong> magnitude of aliasing in computing the real cepstrum of a<br />

speech signal. Your program should compute the real cepstrum of a<br />

frame of speech using an FFT size of 8192 (the st<strong>and</strong>ard) <strong>and</strong> repeat<br />

the computation of the exact same frame of speech with FFT sizes<br />

of 4096, 2048, 1024 <strong>and</strong> 512. Compare the resulting cepstrums <strong>and</strong><br />

plot the differences between the real cepstrum computed using an<br />

8192 point FFT with each of the cepstrums computed using smaller<br />

size FFTs. This exercise illustrates the level of aliasing that is common<br />

in speech processing using reasonable size frames of speech <strong>and</strong><br />

reasonable size FFTs. Use the speech file test 16k.wav to illustrate<br />

the effects of aliasing. Choose a voiced speech frame starting<br />

at sample 13,000, of duration 40 msec for this exercise. Plot the<br />

8192 point FFT cepstrum (the st<strong>and</strong>ard) in the top panel <strong>and</strong> the<br />

differences between the 8192 point FFT cepstrum <strong>and</strong> cepstrums<br />

computed using smaller size FFTs in the lower panel.<br />

8.19 (MATLAB Exercise) Write a MATLAB program to demonstrate<br />

the effects of the cutoff quefrency of the low quefrency lifter on the<br />

smoothing of the log magnitude spectrum. Using a voiced section of<br />

a speech file, calculate the real cepstrum <strong>and</strong> apply a low quefrency<br />

lifter with cutoff quefrencies of 20, 40, 60, 80 <strong>and</strong> 100, <strong>and</strong> show the

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!