4Q Quiz: Short-time Fourier transform (week 4)

Help Center

Warning: The hard deadline has passed. You can attempt it, but you will not get credit for it.
You are welcome to try it as a learning exercise.

In accordance with the Coursera Honor Code, I (Juan Carlos Vega Oliver) certify that the
answers here are my own work.

Question 1
The main relationship between the STFT and the DFT is that
the STFT uses the FFT algorithm and the DFT does not
the STFT is the time-varying version of the DFT
the STFT works for inharmonic signals and the DFT works for harmonic signals
they are basically the same, just the indexes are different

Question 2
The DFT of a windowed sinusoid is
a delta function located at the frequency of the sinusoid
it is a window function centered at 0 Hz
the transform of the sinusoid multiplied by the transform of the window

the transform of the window shifted to the frequency of the sinusoid and scaled with the
amplitude of the sinusoid

Question 3

The analysis windows that we use in this course are

functions that emulate the hearing process
complex and non symmetric functions
rectangular functions centered around 0
real and symmetric functions

Question 4
The main difference between the Blackman window and the Hanning window is that
the length of the Blackman window is bigger than the one of the Hanning window

the spectrum of the Blackman window has a wider main lobe and a lower highest side lobe level
than the Hanning window
they are basically the same, they are just used in different contexts

the spectrum of the Blackman window has a narrower main lobe and a higher highest side lobe
than the Hanning window

Question 5
A bigger window size applied to a stable sound will result into
worst frequency resolution
no effect in the spectral resolution
better time resolution
better frequency resolution

Question 6
The main difference between an even and an odd size window is that
with an odd size window we can use the FFT algorithm, not with an even size window

with an odd size window we can apply zero padding and not with the even size one
with an odd-size window we can apply perfect zero-phase windowing
an even size window is bigger than an odd size one

Question 7
Given a window-size M, if we use an FFT-size N larger than M
we will apply zero padding to the windowed signal
we will get an error because M and N have to be of the same size
we will not be able to compute the FFT because the signal will include zeros
we will have to add more samples of the signal in order to fill the empty space

Question 8
For a perfect reconstruction using the STFT, the window hop-size should be chosen in a way that
we obtain an amplitude modulation in the resulting sound
the overlapping of the windows is equal to a constant
the windows are overlapped by 50% of their window length
we have an even number of frames in the analysis

Question 9
The time-frequency compromise of the STFT refers to the fact that
the more time resolution we get, the more frequency resolution we also get
it is not possible to control the trade-off between time and frequency resolution
we can only obtain frequency resolution, we will never be able to improve the time resolution

if we want to improve the frequency resolution by increasing the window size we lose time
resolution

Question 10
A magnitude spectrogram is
the visualisation of the sequence of magnitude spectra obtained using the STFT
the combination of magnitude and phase spectrum of the DFT
the visualisation of the sequence of phase spectra obtained using the STFT
the magnitude spectrum resulting from the DFT

In accordance with the Coursera Honor Code, I (Juan Carlos Vega Oliver) certify that the
answers here are my own work.

Submit Answers

Save Answers

You cannot submit your work until you agree to the Honor Code. Thanks!