XE31CZS Exercise - Periodicity in Short-Time DFT Spectrum
Tasks to do:
- Simulated and Real Periodicity in DFT
- Load one period of voiced speech signal vm0-per.bin.
(The signal is binary, to load it into MATLAB
use attached function loadbin.m )
and find shortime DFT of these signal. Lenght of the DFT should be
same as the signal (period) length.
- Observe the changes in the spectrum when the given signal is
extended by zeros into 2, 3, 4 -times higher length.
- Observe the changes in the spectrum when the length of the signal
is changed by repetition of the same period. Realize it again for 2,
3, 4 -times repetition of the period.
- Observe the changes in the spectrum when the signal with 4 times
repeated period is analyzed by DFT of the length
a) 479, 481
b) 470, 490
c) 256, 512
- Observe the periodicity in real signal (voiced sound of speech) vm0.bin, evaluate it for lengths 128, 256, 512, 1024.
- Estimate fundamental period (and fundamental frequency) of sounds
vf5.bin and vm5.bin from their
short-time DFT spectra.
- Observe short-time DFT spectra for unvoiced sound of speech uf5.bin and um5.bin
- Weigthing and the Periodicity in ST-DFT
- Observe periodograms (short-time DFT spectrum) of above available
signals for widow lengths 128, 256, 512, 1024.
- Observe and explain the changes in achieved results when different
weighting windows are used (rectangular, Hamming, Blackman).
- Evaluation of convolution by DFT
- Create input traingular sequence x1[n] of length 200
- Create input rectangular sequence x2[n] of length 200
- Evaluate convolution of these two sequences using DFT for N =
200, 400, 256, 500 samples
NOTE. Compare linear and
logaritmic scale of y-axis for observation of short-time magnitude spectra (i.e.
'plot(f,abs(S))' or 'semilogy(f,abs(S))' respectively).