How identify echo patterns in audio samples

Question

How do I estimate if in recorded sound samples is present an echo caused from walls reflection ? (maybe there are multiples echoes: delayed, attenuated versions of the signal itself).

I'm recording several sound samples in my room and I suppose that in the recorded audio samples is present an echo image of the original sound. How do I verify if this is true and in case, how can I calculate the samples delay (ms echo reflection time from the walls) ?

I plot the cross-correlation of the samples with itself (xcorr in MATLAB, about 40000 samples, ~1 second) but I don't know which pattern to look for to check if one or more echoes are present.

If I can understand the presence of the max peak in the middle of the graph, what are all the others equally spaced peaks on the both sides of the xcorr plot ? Eventually, how do I estimate the echo parameters from the recorderd samples (ms delay or samples delay) ?

Answer 1

The side peaks could be the result of periodic structure in your input signal (say around on the order of 8-10Hz, plus a 2nd harmonic).

If you are creating the sound samples, try playing pseudo-random noise, which has good autocorrelation properties for a long enough sample, notably tight peaks & aperiodic.

2nd John's answer. You can do a back-of-the-envelope estimate of the max echo lag (assume a few reflections) using the speed of sound & room size.

Answer 2

Try limiting the max lag of the xcorr based on the largest delay that you care about.