If I have two different signals, say of the same length and the sampling frequency, but these two signals do not have significant differences in their frequency spectrum. I wanted to ask you about the available methods of identifying the features that can quantitatively distinguish the two signals.
1.may you can try the coherence function i.e. x.y/norm(x)*norm(y)
2. the time-frequncey methods also help you. because the two signals in the same distriubtion in frequency domain can be distinguished in time-frequncy domain.
performing a histogram difference (ie separate the signal in frequency bands, then subtract their magnitudes in each band, make sure you normalize it though) is a simple technique that might be useful in this case.
Dear thaneshwor,
you can find the function of your graph or signal and then put it into matlab software and then use it's graphic options in order to find out even detailed differences of those signals or graphs easily.
good luck...
Assuming your two signals have the same spectrum, i.e. the same modulus of their Fourier transforms, the signals are different only if the phases of the FT are different. So computing the two phases their difference, or any feature computed from this difference will be a candidate to show the two signals are different.
If the spectra are (almost) the same, then the signals are (almost) the same. But if you mean the amplitude spectrum only, then there might be differences in the phase, as it was rightly pointed out by Hervé Rix.
Notice also that, in general, the amplitude spectrum is related only to the mean frequency characteristics of the signal (averaged over the duration of the signal). Different frequency components might occur in both signals in different time sections, which will be evident in the phase spectrum and which might be more straightforwardly visualized via wavelet analysis.
Besides, you might drop frequency-domain analysis of the signal and perform your comparison entirely in time domain (correlation, autocorrelation functions, different norms, etc.), as it was suggested by Baijie Qiao.
- Badges
- Science topic