search

IoSR-Surrey / MatlabToolbox

General purpose Matlab toolbox
MIT License
128 stars 43 forks source link

Regarding the delay introduced by octave band filterbank #9

Closed ghamaa closed 6 years ago

ghamaa commented 6 years ago

Hello, In the irStats script you decomposed the room impulse response into octave bands. Then, you compute the reverberation time RT in each band. My issue is regarding the time delay introduced by each filter, should we compensate for that delay when we compute the reverberation time in each band. Thank you for your help.

chummersone commented 6 years ago

No. The function automatically determines the start point in each band as the peak of the squared impulse response.

ghamaa commented 6 years ago

You computed the peak with respect to the full band room impulse response (RIR) as stated in irStats: peak = find(x(:,n).^2==max(x(:,n).^2)); If you filter RIR and there is a time delay introduced by the filter then the location of the peak will be different from that of the full band RIR.

chummersone commented 6 years ago

The octave-band decay measures don’t use the position of the initial peak directly. They measure the time separating the fit points.

ghamaa commented 6 years ago

You are correct if we assume linear phase of filters within their passband. What about the frequency dependent direct-to-reverberation ratio (DRR). It depends on the position of the initial peak in each frequency band.

chummersone commented 6 years ago

ISO 3382 makes certain requirements of the octave-band filters. The filters were designed to meet those requirements.

DRR is not calculated in each octave band, only for the entire response, per the ISO 3382 standard.