MATLAB: How to use pspectrum() function

Question

Hello,

I want to create a power spectrum vs frequency plot of an audio signal I have. I tried using the following code:

[speech, f1] = audioread('outputTSNRcut.wav');
[noise, f2] = audioread('input2noise.wav');
[speechInput,fspeech] = pspectrum(speech, f1);
[noiseInput, fnoise] = pspectrum(noise, f2);
fig1 = figure;
plot(fspeech, abs(speechInput), 'b');
xlabel('Frequency (Hz)');
ylabel('Power Spectrum (dB)');
title('Speech Spectral Density');
fig2 = figure;
plot(fnoise, abs(noiseInput), 'r');
xlabel('Frequency (Hz)');
ylabel('Power Spectrum (dB)');
title('Noise Spectral Density');

which produced the following power spectrum plot:

The problem is that the audio is a speech recording, thus I'm expecting the power spectrum to be somwhere around 20 – 40 dB, perhaps more, but certintly not under 1 dB! I'm not sure what have I done wrong, I tried to follow exactly the pspectrum documantation. I'd appreciate it if someone could correct me.

Thank you.

Answer 1

Hi Elinor, the ppsptectrum function does not return the output directly in dB to be plotted. According to the documentation,

For the plot, the function converts p to dB using 10*log10(p)

• You can use ppspectrum with no output arguments to plot the spectral estimate in the current window.

        ppspectrum(speech, f1);

• Another way is to manually convert p to dB.

Hope this helps.