From: Eric Durant <edurant@umich.edu>
Subject: Re: Simplified HRTF model
Date: 12 Oct 2000 00:00:00 GMT
Message-ID: <39E543B3.A6449CCC@umich.edu>
Content-Transfer-Encoding: 7bit
References: <8s25sa$f2t$1@flood.xnet.com>
X-Accept-Language: en
Content-Type: text/plain; charset=us-ascii
X-Complaints-To: abuse@earthlink.net
X-Trace: newsread1.prod.itd.earthlink.net 971326437 63.11.55.247 (Wed, 11 Oct 2000 21:53:57
PDT)
Organization: University of Michigan
MIME-Version: 1.0
NNTP-Posting-Date: Wed, 11 Oct 2000 21:53:57 PDT
Newsgroups: comp.dsp,comp.games.development.audio

It sounds like you'd do well to do a principal components
analysis (PCA) of your HRTF magnitude data and then mix the
minimum phase impulse responses of the first five according to
their position dependent projections.  See:

Doris J. Kistler and Frederic L. Wightman, ``A model of
head-related transfer functions based on principal components
analysis and minimum-phase reconstruction,'' The Journal of the
Acoustical Society of America (JASA), vol. 91, no. 3, pp.
1637--1647, 1992.

There are also low complexity IIR approximation methods, but it
sounds like you've ruled these out for your hardware.

-- Eric Durant
   http://edurant.com/

Snowfox wrote:
>I'm looking for a simplified model for head related
>transform functions.
> 
>Specifically, I am implementing an advanced 3D library on a
>platform with a large number of hardware playback channels
>but no significant sample processing capabilities.
> 
>I would like to accomplish an approximation of HRTFs by
>blending preprocessed samples.