Tsourouflas
Newbie level 2
Hi all, i am new to this site and i would like your help on a matter.
i am calculating the SDM (self distance matrix) of a song represented by its chromagram. since the SDM is symmetric i continue with the lower left triangle of it. that far all good.
next step is i want to detect line segments for each diagonal .
i calculate the sum for each diagonal and put it in F(k), so that F(1) corresponds to the sum of the 1st diagonal f(2) the second and so on..
next step is 'removing cumulative noise' and here is where i need your help:
-> im told i have to "detrend" F(k) to remove cumulative noise
-> it should be achieved by calculating a lowpass filtered version of F(k) using a FIR filter with 50 taps, the value of each coefficient being 1/50.
-> next the lowpass filtered version of F(k) is subtracted from F(k)
i saw the other threads about fir filters but it didnt help me that much.
any suggestions, explanations, code on this matter is highly appreciated.
i thank you in advance,
nik
i am calculating the SDM (self distance matrix) of a song represented by its chromagram. since the SDM is symmetric i continue with the lower left triangle of it. that far all good.
next step is i want to detect line segments for each diagonal .
i calculate the sum for each diagonal and put it in F(k), so that F(1) corresponds to the sum of the 1st diagonal f(2) the second and so on..
next step is 'removing cumulative noise' and here is where i need your help:
-> im told i have to "detrend" F(k) to remove cumulative noise
-> it should be achieved by calculating a lowpass filtered version of F(k) using a FIR filter with 50 taps, the value of each coefficient being 1/50.
-> next the lowpass filtered version of F(k) is subtracted from F(k)
i saw the other threads about fir filters but it didnt help me that much.
any suggestions, explanations, code on this matter is highly appreciated.
i thank you in advance,
nik