# Analysis of two audio signal by autocorrelation

I have recorded my audio by two hardware, PC and Microcontroller, I know the PC result are correct but my signal have some deficit. I have recorded by my PC and MCU my voice 6s 7812Hz and autocorrelation of each one become : Auto in case excitation being white noise: It seems in MCU we have summed with triangle signal in comparison with the PC signal autocorrelation.

Crosscorrelation :

White noise excitation cross-correlation (note both signal are recorded from one noise with small delay(PC have delay)):

I know what autocorrelation function is. Why mcu auto-corr has summed with triangle? Also two signal must be at some point highly correlated but its weird! what can be inferred?

## Update

Why cross correlation of white noises is not spike? Also note the end of cross being zero, it have zero tail, why?

The "MCU" plot looks like there was a DC component in the signal, that ran through the autocorrelation function. The triangle is typical for this. Try highpass filtering before the analysis.

Regarding your second question: the plot looks quite normal for autocorrelation of a speech signal. There are always tonal/peridodic components in speech, which is why sometimes the maxium of the autocorrelation function isn't quite clear. In fact, this plot looks like there could be a repetition of the same word or syllable in the speech material.

EDIT:

The small plots of whitenoise autocorrelation are looking good, one clear peak, as can be expected.

The big plot seems to be the crosscorrelation (it also says so in the text) of two different white noise signals. There can't be any peak, as the signals are in no way correlated.

• How you've understand there is a repetition of syllable or word? because cross correlation is broadband? Commented May 30, 2022 at 8:59
• The three very distinct peaks are telltale for a repetition of some kind.
– Max
Commented May 30, 2022 at 9:00
• Thank you so much. yes I'm using GNU octave xcorr function. Commented May 30, 2022 at 9:03
• Could you please watching the question again since I've updated it with white noise data. Commented May 30, 2022 at 14:20
• I've made an edit to the answer.
– Max
Commented May 30, 2022 at 14:25