Signal Processing / Model Evidence !!

Hello everyone.

I am just wondering if there is an algorithm or technique to give the answer to the question "If the two signals are originally from the same source".. I am trying to calculate the delay time between two signals and the data are real data so I don't know my answer myself but I meant to calculate the delay time then be able to answer if the two signals are alike but just that one is delayed but at the moment with me I just answer this question by looking at the signals after shifting it back by the delayed time. My question is, is there a more practical way to code then it will be like a model evidence rather than my own judgement based on some alogorithm it will answer this question?? Any idea on any algorithm I should use and read about will be appreciated !!

Cheers, Susan

[Fangjun Jiang]

I am trying to calculate the delay time between two signals and the data are real data so I don't know my answer myself but I meant to calculate the delay time then be able to answer if the two signals are alike but just that one is delayed but at the moment with me I just answer this question by looking at the signals after shifting it back by the delayed time.

breathe, breathe, breathe...

Answer 1

[David Young]

In general, you need to make some assumptions about what might have happened to the signals (e.g. addition of noise, non-linear distortion, delays etc.) in order to answer your question. However, it sounds like you believe that there are two possibilities: (a) the signals are unrelated or (b) the main difference is just a time delay. In that case, cross-correlation may be sufficient. Start by having a look at that.

[Susan]

Thanks for your answer, I did use cross correlation to calculate the delay time between the two signals and yeah in my case I am not looking at any other factors other than it could be delayed or they not from the same source. but when I run the code it calculates the delay time then I need to shift it by the delayed i calculated then judge it myself by looking at it if they are actually from the same source but one is delayed, I am asking is there a method or technique I could implement that give me the answer by using the delayed time was calculated and determine if the indeed the same but one is delayed or they absolutely unrelated rather than my own judgement?

[David Young]

The cross correlation gives you a correlation coefficient, as well as the best lag. The correlation coefficient will be larger the more similar the two signals are. You can do a statistical test on it (if you can make suitable assumptions about the noise distribution) to help make a judgement about whether the signals are from the same source - try a web search on "correlation coefficient statistical test".

[Susan]

Thank you, will look at it !!

[Daniel Shub]

It is a minor, but important point, most statistical tests will tell you how confident you are that the signals are NOT from the same source. They tell you little about if the signals are in fact from the same source.

[Susan]

Thank you Daniel, I managed to get the ttest2 works on my code and you right it tells me that the two signals are not from the same source but do you know what to use to find out if they are from the same source??

Answer 2

[Syeed]

Hello everyone,

In Signal processing onrap course there is an task in Preprocessing Signals of Custom Preprocessing Function and the task is "Duplicate the harp, pax, and wanc signals and append _cm to the signal names. For example, duplicate harp and name the result harp_cm.

The function nm2cm has been written for you. Use the custom preprocessing function nm2cm to process the three new signals so that their units are in centimeters instead of nanometers. Then display the three new signals in the time domain." and I put my 100% on it. so, can you gays help me.

[Ayush]

Hello

You need to preprocess all the three copies with the nm2cm function, one by one. This can be verified by the 'i' button which will appear after you have applied the function to the copy.

Hope this helps!!

[CHEEPATI]

me also put 100% effort can't solve it can you explain in detail.

[Abhishek Tripathi]

Duplication was succesful but it says the harp_cm, pax_cm, and wanc_cm signals are not processed correctly.

Answer 3

[Vasam]

Duplicate the harp, pax, and wanc signals and append _cm to the signal names. For example, duplicate harp and name the result harp_cm.

The function nm2cm has been written for you. Use the custom preprocessing function nm2cm to process the three new signals so that their units are in centimeters instead of nanometers. Then display the three new signals in the time domain.