TY - JOUR
T1 - Optimal Series Truncation of the Rigid-Sphere Head-Related Transfer Function for Accurate Binaural Perception
AU - Sridhar, Rahulram
AU - Choueiri, Edgar Y.
N1 - Funding Information:
This work was jointly sponsored by Focal-JMlab and Tesla, Inc. The authors wish to thank Dr. Joseph G. Tylka for fruitful discussions and reviewing an early draft of the manuscript and the anonymous reviewers for their valuable feedback during the peer-review process. All plots in this paper were generated in Mathematica using the open-source MaTeX software [43] for LaTeX typesetting.
Publisher Copyright:
© 2021 Audio Engineering Society. All rights reserved.
PY - 2021/5
Y1 - 2021/5
N2 - A formula is derived for computing the order at which the infinite series representation of the rigid-sphere head-related transfer function (RS-HRTF) must be truncated to minimize the time required to compute the HRTF to a sufficiently high accuracy based on binaural perception metrics. Quick and accurate computation of this HRTF may be useful for implementing spatial audio in computationally limited and portable devices. Using a brute-force approach, the lowest truncation order, Nmin, that yields the RS-HRTF that differs from the benchmark (i.e., the RS-HRTF computed with the highest possible accuracy) by less than just-noticeable difference thresholds in interaural time and level differences is approximately computed for a wide range of source distances. By fitting power and rational functions to these computed values, a formula that approximates Nmin as a function of frequency and source distance is derived. It is shown that truncation order varies significantly with source distance and that the proposed formula, unlike a previous one, accurately captures this variation. Consequently, using the proposed formula instead of the previous one results in a more accurate RS-HRTF that is also computed 48% faster on average.
AB - A formula is derived for computing the order at which the infinite series representation of the rigid-sphere head-related transfer function (RS-HRTF) must be truncated to minimize the time required to compute the HRTF to a sufficiently high accuracy based on binaural perception metrics. Quick and accurate computation of this HRTF may be useful for implementing spatial audio in computationally limited and portable devices. Using a brute-force approach, the lowest truncation order, Nmin, that yields the RS-HRTF that differs from the benchmark (i.e., the RS-HRTF computed with the highest possible accuracy) by less than just-noticeable difference thresholds in interaural time and level differences is approximately computed for a wide range of source distances. By fitting power and rational functions to these computed values, a formula that approximates Nmin as a function of frequency and source distance is derived. It is shown that truncation order varies significantly with source distance and that the proposed formula, unlike a previous one, accurately captures this variation. Consequently, using the proposed formula instead of the previous one results in a more accurate RS-HRTF that is also computed 48% faster on average.
UR - http://www.scopus.com/inward/record.url?scp=85111103619&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85111103619&partnerID=8YFLogxK
U2 - 10.17743/JAES.2021.0007
DO - 10.17743/JAES.2021.0007
M3 - Article
AN - SCOPUS:85111103619
SN - 1549-4950
VL - 69
SP - 340
EP - 350
JO - AES: Journal of the Audio Engineering Society
JF - AES: Journal of the Audio Engineering Society
IS - 5
ER -