MIT KEMAR HRTF DATA

[Bill Kapralos]

Hello!

Perhaps someone can help me out with a problem/question I have regarding
the use of the MIT KEMAR HRTF dataset. I am using the compact dataset
(128 samples per HRTF) and have no problem reading in the data etc.
However, the resulting spatialized sound is very poor - e.g. sound is not
spatialized to the location corresponding to the HRTF but rather seems
like it is coming from behind and the level of the spatialized sound
varies form location to location for example take a constant elevation and
take two hrtfs with differeng azimuth (e.g. by 5 degrees) and the level
may differ drastically)... The only processing I have performed on the
data is to switch the raw data to little endian (I am using an Intel
based PC) and to normalize the samples of ecah measurement by the maximum
sample value of the entire dataset (e.g. all samples to fall between 1 and
-1). I am also determining (and applying) the appropriate time delay and
my un-processed sound is also sampled at 44.1kHz as with the hrtf data.

This is my first time using the dataset and I have heard that results are
not too great generally however I would like to hear any
comments/suggestions!

Thanks in advance!

Please reply to billkATcs.yorku.ca

[j***@gmail.com]

I wish I could offer some advice, but instead I have a question. How can I gain access to the MIT KEMAR non-individualized HRTFs. I have googled them and looked on the MIT website to no avail.

Jeremy G. Gonzalez

Post by Bill Kapralos
Hello!
Perhaps someone can help me out with a problem/question I have regarding
the use of the MIT KEMAR HRTF dataset. I am using the compact dataset
(128 samples per HRTF) and have no problem reading in the data etc.
However, the resulting spatialized sound is very poor - e.g. sound is not
spatialized to the location corresponding to the HRTF but rather seems
like it is coming from behind and the level of the spatialized sound
varies form location to location for example take a constant elevation and
take two hrtfs with differeng azimuth (e.g. by 5 degrees) and the level
may differ drastically)... The only processing I have performed on the
data is to switch the raw data to little endian (I am using an Intel
based PC) and to normalize the samples of ecah measurement by the maximum
sample value of the entire dataset (e.g. all samples to fall between 1 and
-1). I am also determining (and applying) the appropriate time delay and
my un-processed sound is also sampled at 44.1kHz as with the hrtf data.
This is my first time using the dataset and I have heard that results are
not too great generally however I would like to hear any
comments/suggestions!
Thanks in advance!
Please reply to billkATcs.yorku.ca