Why is there low frequency response attenuation on HRTFs (or impulse responses) that are windowed to a shorter length. One explanation I've heard is that because the measurements are AC- coupled having less bins on the FFT "smooths" the frequency response. I would like to be able to calculate this attenuation for a window of any size. Any clues on how to do this? I haven't found any literature on the subject.... --------------------------------------- Posted through http://www.DSPRelated.com
Loss of low frequency response on a windowed HRTF
Started by ●September 28, 2015
Reply by ●September 29, 20152015-09-29
If the impulse response has long low-frequency tails that exceed your window length then you will see a reduction in low frequencies. Given the physical dimensions of the outer ear and ear canal, it's unlikely that the HRTF curves will have any features below 500 Hz or so. Since the measurement of the HRTF involves both a loudspeaker and the surrounding room acoustics (if any), then it might be the case that any low-frequency response variations are from the test setup rather than the actual acoustics of the outer ear/ear canal, and therefore you should focus on improving that aspect of your measurement. Alternatively, you can measure the speaker-to-microphone path with no ear/head present, and then de-convolve the effects of the speaker + acoustic path. But this can prove to be a bit messy in practice. Another alternative to to fill in the low frequencies by repeating the lowest fft bin that you "trust". This approach is validated by the assumption above, that nothing "interesting" can be happening at wavelengths that are far larger than the physical dimensions of the thing you are measuring. Bob On Monday, September 28, 2015 at 1:42:46 PM UTC-4, alfredoFranco wrote:> Why is there low frequency response attenuation on HRTFs (or impulse > responses) that are windowed to a shorter length. > > One explanation I've heard is that because the measurements are AC- > coupled having less bins on the FFT "smooths" the frequency response. > > I would like to be able to calculate this attenuation for a window of any > size. Any clues on how to do this? I haven't found any literature on the > subject.... > > > --------------------------------------- > Posted through http://www.DSPRelated.com
