Browse Prior Art Database

Apparatus and Methods for Efficient Deployment of Head-Related Transfer Functions

IP.com Disclosure Number: IPCOM000126020D
Publication Date: 2005-Jun-28

Publishing Venue

The IP.com Prior Art Database

Abstract

The use of all-pole filters for the deployment of head-related transfer functions (HRTFs) is considered, with the purpose being the generation of binaural (3D) audio signals. All-pole filters are members of a general class of digital filters known as infinite impulse response (IIR) filters. The advantages of all-pole filters lie in their simplicity in design, efficiency in implementation, and straightforward interpolation among filters. A novel neural network- based parameter storage device is introduced to work with all-pole filters; the device provides great saving in storage requirements, as well as direct interpolation to the underlying parameters; allowing therefore the recovery of filter parameters for continually-varying source coordinate values. Several embodiments utilizing the explained principles are described, which achieve significant improvement in efficiency with respect to prior art by reducing computational complexity and storage requirement.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 10% of the total text.

Page 1 of 20

DoCoMo Communications Laboratories USA, Inc.

181 Metro Drive Suite 300, San Jose, CA 95110, USA

1 TITLE OF THE INVENTION

Apparatus and Methods for Efficient Deployment of Head-Related Transfer Functions

2 INVENTOR

Wai C. Chu

3 ABSTRACT

The use of all-pole filters for the deployment of head-related transfer functions (HRTFs) is considered, with the purpose being the generation of binaural (3D) audio signals. All-pole filters are members of a general class of digital filters known as infinite impulse response (IIR) filters. The advantages of all-pole filters lie in their simplicity in design, efficiency in implementation, and straightforward interpolation among filters. A novel neural network- based parameter storage device is introduced to work with all-pole filters; the device provides great saving in storage requirements, as well as direct interpolation to the underlying parameters; allowing therefore the recovery of filter parameters for continually-varying source coordinate values. Several embodiments utilizing the explained principles are described, which achieve significant improvement in efficiency with respect to prior art by reducing computationa l complexity and storage requirement.

4 BACKGROUND OF THE INVENTION

The spectral filtering of a sound source before it reaches the ear drum that is caused primarily by the outer ear is termed the head-related transfer function (HRTF), which captures the sound localization information associated with incident sound waves [21]. In practice the HRTFs are often measured from live subjects or dummy heads by playing a probe signal at a given position and measuring the impulse response with microphones placed in the vicinity of the ear canals; the resultant signals are called head-related impulse response (HRIR) and can be directly used in signal filtering so as to synthesize binaural sound from a monaural source.

  In the language of signal processing, the direct use of HRIR is known as finite impulse response (FIR) filtering or all-zero filter, since the associated system function only contains zeros. The major problem with FIR filters is computational complexity: the impulse response is relatively long with a large number of samples, leading to elevated cost in storage and computation. Moreover, interpolation between filters cannot be trivially done [21, p.166], which is undesirable for head-tracked applications, where the perceived change in direction must be smooth. Figure 1 shows the block diagram of

1

[This page contains 1 picture or other non-text object]

Page 2 of 20

DoCoMo Communications Laboratories USA, Inc.

181 Metro Drive Suite 300, San Jose, CA 95110, USA

a system where an input monaural signal is converted to binaural output bearing localization information. Note in Figure 1 that the direction of the source, or the source coordinate, is specified by two angles: azimuth and elevation corresponding to the interaural polar coordinate system [29].

Figure 1. Generation of 3D au...