Signal Analyzer

Load a speech signal sampled at $$F_s=7418Hz$$. The file contains a recording of a female voice saying the word "MATLAB®."

load mtlb

To simulate a situation where 70% of the audio data is missing, randomly assign NaN values to the signal.

rng(2024) 
numToReplace = round(length(mtlb) * 0.70);
missing = randperm(length(mtlb),numToReplace);

mtlbMissing = mtlb;
mtlbMissing(missing) = NaN;

Open Signal Analyzer and drag the mtlb and mtlbMissing variables from the Workspace Browser pane to the Filter Signals table. Select the two signals. On the Analyzer tab, click Time Values and select Sample Rate and Start Time. Specify Sample Rate as Fs Hz and Start Time as 0 s. Click Display Grid to create two side-by-side displays. Plot mtlb in the left display and mtlbMissing in the right display. To hear the mtlb audio signal, select it and click Play in the Playback section of the toolstrip under the Display tab. To repeat the signal, select Play in Loop before playing.

Select the signal with missing data and click Preprocess under the Analyzer tab to enter the preprocessing mode, then select Fill Missing from the list of preprocessing options. Use the Function Parameters panel to adjust the Fill Missing parameters. Select Autoregressive model and click Apply to fill in the missing signal. Click Accept All to save the preprocessing results and exit the mode. For details on alternative fill functions, see fillmissing and fillgaps.

You can now play the filled signal using the Play button. To see the effect of filling missing data on the spectrogram, click Time-Frequency on the Display tab. On the Spectrogram tab, specify a time resolution of 20 ms and 80% overlap between adjoining segments. Set the Power Limits to –50 dB and –10 dB. Click the left display and repeat the steps.

You can adjust the spectral leakage of the analysis window to resolve sinusoids in Signal Analyzer.

Generate a two-channel signal sampled at 100 Hz for 2 seconds.

fs = 100;
t = (0:1/fs:2-1/fs)';

x = sin(2*pi*[20 20].*t)+[1 1/100].*sin(2*pi*[21 30].*t);

Embed the signal in white noise. Specify a signal-to-noise ratio of 40 dB.

x = x + randn(size(x)).*std(x)/db2mag(40);

Open Signal Analyzer and plot the signal. On the Analyzer tab, with the signal selected in the Signal table, click Time Values and select Sample Rate and Start Time. Specify Sample Rate as fs Hz and Start Time as 0 s. On the Display tab, click Spectrum to add a spectral plot to the display.

Click the Spectrum tab. The slider that controls the spectral leakage is in the middle position, corresponding to a resolution bandwidth of about 1.28 Hz. The two tones in the first channel are not resolved. The 30 Hz tone in the second channel is visible, despite being much weaker than the other one.

Increase the leakage so that the resolution bandwidth is approximately 0.83 Hz. The weak tone in the second channel is clearly resolved.

Move the slider to the maximum value. The resolution bandwidth is approximately 0.5 Hz. The two tones in the first channel are resolved. The weak tone in the second channel is masked by the large window sidelobes.

Click the Display tab. Use the horizontal zoom to magnify the frequency axis. Add two cursors to the display and drag the frequency-domain cursors to estimate the frequencies of the tones.

Read an audio recording of an electronic toothbrush into MATLAB®. The signal is sampled at 48 kHz. The toothbrush turns on at about 1.75 seconds and stays on for approximately 2 seconds.

[y,fs] = audioread("toothbrush.m4a");

Open Signal Analyzer and drag the signal from the Workspace Browser to the Signal table. Add time information to the signal by selecting it in the Signal table and clicking Time Values on the Analyzer tab. Select Sample Rate and Start Time and enter fs for the sample rate.

On the Display tab, click Display Grid to create a two-by-two grid of displays. Select each display, click Spectrum to add a spectrum view, and click Time to remove the time view. Drag the signal to all four displays.

Click the Spectrum tab to modify the spectrum view in each display.

You can see that some views show higher resolution but higher leakage, while other views have lower leakage but at the expense of resolution.