0:34

So well start with the sonic visualizer.

Â And this is the sound I want to analyze,

Â it's a note of a soprano singing and let's hear that.

Â [SOUND] Okay, that's a quite high pitch of sound, and

Â it has quite a bit of a vibrato, this frequency oscillation

Â that is characteristic of operatic singing.

Â And to understand a little bit better the sound, let's open the spectrogram,

Â so let's open the pane on the spectrogram.

Â So here it is, and so we see now clearly some information of this voice sound,

Â we see these horizontal lines that correspond to the harmonics,

Â and we see this oscillation, which is basically this vibrato that is present.

Â To go a little bit deeper, let's open another pane with

Â a single spectrum of this time-veering spectrogram.

Â So let's first show it as in a linear scale, the horizontal axis.

Â Let's have lines as interpolation.

Â And the window let's use the same window done for the STFT.

Â So we'll use 1024.

Â Okay, and this is it.

Â Let's maybe make it zoom in and stretch so

Â that we see same things that we would see in the spectrogram, okay?

Â So this is one slice of the spectrogram, so

Â all these horizontal lines correspond to these peaks that we see here, okay?

Â And now maybe let's change things.

Â Let's change for example the window size.

Â If we change to 256 both analysis, okay?

Â Now, what we're seeing is a much smoother shape,

Â we are not seeing the individual harmonics, we're just seeing

Â an overall shape which basically correspond to what we call the formants.

Â The resonances of the vocal track which is what makes us be

Â able to distinguish between vowels for example.

Â So each vowel has a characteristic formant structure.

Â Of course if we move, things will change a little bit.

Â But of course the vowel remains the same so it will not change that much.

Â 3:04

If we go back to the analysis size that

Â allows us to visualize the harmonics and when we move well, we see few more changes

Â because the harmonics are changing more than just the formants, okay?

Â So now if we want to change the type of window,

Â we can go to the Preferences and

Â in the Analysis tab, let's put this away from this.

Â Here in the last option, we see the analysis window to be used.

Â Okay, curly is the blackman window.

Â Let's change for example to rectangular window,

Â which is the square that would cut the signal very abruptly.

Â Let's apply that.

Â Okay, so clearly, it doesn't look so nice.

Â 5:00

let's use the same values that we use before, 1024 for

Â both FFT size and window size with the blackman window.

Â And we can compute, okay, and this is basically what we saw before.

Â This is the 1024 samples we have started with, this is the magnitude and

Â phase spectrum and we see clearly the peaks corresponding to the harmonics.

Â I mean, here we see the phase spectrum, which we didn't see before, and

Â also we see the inverse of this.

Â So, this is the windowed signal that we generate back

Â by taking the inverse FFT of this spectrum.

Â And, of course, we can do the same thing.

Â We can change windows, for example,

Â if we change the window size to 256, and also the 50 size to 56.

Â In the same location, we compute, well,

Â we see we are, of course, taking much less samples and

Â the spectra are much smoother, less information there, okay?

Â One advantage, of course, with this interface we have is that we

Â can independently control the window size from the FFT size.

Â So we can put FFT size 1024 and

Â maybe a window size not that large, maybe 801.

Â It will compute, okay?

Â We are taking less samples than before but

Â still the frequency resolution is quite good.

Â And it's quite smooth because we have been doing zero padding and

Â so the shape of the spectrum is quite nice.

Â Okay, let's look at all these from the short-time Fourier transform

Â perspective from the spectogram.

Â So we will get the same sound.

Â Okay, and again, let's put 1024, 1024 and

Â the hop size has to be at least much smaller than the window

Â size in a way that they overlap at factor as correctly.

Â So for 1024 in the blackman window, at least we need one-fourth.

Â So let's put 256 and we compute.

Â Okay, and this is the result.

Â So we have the input signal.

Â The magnitude spectrogram, the phases

Â of the time-varying phases and the output sound and the output sound.

Â [SOUND] Well, it's very much the same than the original because we

Â have done a good reconstruction with a good overlap.

Â However, if this overlap is not correctly set for

Â example, let's put the same than the windows size for

Â example, 1024 and let's compute it.

Â Well, clearly now is something wrong in the output signal and

Â if we can listen to it [SOUND], okay?

Â Of course, we see this modulation that is at the frame rate because

Â we are not overlapping correctly, so every frame we see a burst of sound,

Â and they don't balance out by the overlap factor.

Â So we definitely need to have a much smaller hop size.

Â 8:35

And anyway, that's all I wanted to say.

Â Basically, I encourage you to play around with these parameters.

Â You can change the window size, you can change the FFT size, the hop size,

Â the type of window.

Â And in between the DFT and the STFT,

Â I believe you can get a good grasp of all these different parameters and

Â the effect they have in the visualization of the spectrogram and

Â also in the reconstruction of the signal.

Â So, let's just finish and okay, today basically,

Â we have used SonicVisualizer to analyze voice sound and

Â to visualize the spectrogram and the individual spectrum.

Â We have also done the same thing with the interface of the sms-tools.

Â And of course we have used the sound is soprano sound from freesound.

Â 9:43

and has allowed you to understand how useful might be to use these type of

Â techniques to visualize a particular sound, in this case, a soprano sound.

Â Of course this is just the beginning of this more complex analysis.

Â So in next demonstration class, we're going to analyze a more complex sound.

Â And we will see how we can analyze time-varying sound,

Â that have much more structure and

Â how we can use this spectrogram analysis to get some insights on that.

Â So thank you very much for your attention and I hope to see you next class.

Â