A nonlinear personalized model is established to predict individual head-related transfer function (HRTF). Principal component analysis (PCA) is applied to obtain individual weights as the outputs of the model. Some important anthropometric parameters are selected according to the Laplacian score and correlation analyses between all measured parameters and the individual weights. They act as the inputs to the model. Constructing a radial basis function (RBF) neural network to learn the nonlinear model, the individual HRTF can be predicted according to the measured anthropometric parameters. Simulation results show that the proposed method outperforms the partial least squares regression (PLSR) method in predicting individual HRTF.
HUANG Qing-hua, LI Lin, LAI Shi-cun
. Personalized Modeling of Head-Related Transfer Function Based on RBF Neural Network[J]. Journal of Shanghai University, 2014
, 20(2)
: 157
-164
.
DOI: 10.3969/j.issn.1007-2861.2013.07.019
[1] 谢菠蒜. 头相关传输函数与虚拟昕觉重放 [M]. 北京: 国防工业出版社, 2009: 1268-1285.
[2] Xie B, Zhong X, Rao D, et al. Head-related transfer function database and its analyses [J]. Sci China: Ser G, 2007, 50: 267-280.
[3] Blauert J, Brueggen M, Bronkhorst A W, et al. The AUDIS catalog of human HRTFs [J]. Journal of the Acoustical Society of America, 1998, 103(5): 2901-2902.
[4] Zeng X, Wang S, Gao L. A hybrid algorithm for selecting head-related transfer function based on similarity of anthropometric structures [J]. Journal of Sound and Vibration, 2010, 329(19): 4093-4106.
[5] Middlebrooks J, Duda R. Individual difference in external ear transfer functions reduced by scaling in frequency [J]. Journal of the Acoustical Society of America, 1999, 106(3): 1480-1492.
[6] Algazi V, Duda R, Duraiswami R, et al. Approximating the head-related transfer function using simple geometric models of the head and torso [J]. Journal of the Acoustical Society of America, 2002, 112(5): 2053-2064.
[7] Nishino T, Inoue N, Takeda K, et al. Estimation of HRTFs on the horizontal plane using physical features [J]. Applied Acoustics, 2007, 68(8): 897-908.
[8] 汤永清, 方勇, 黄青华. 一种局部线性嵌入的空间昕觉重建方法 [J]. 上海大学学报: 自然科学版, 2011, 17(1): 119-224.
[9] Xie B. Recovery of individual head-related transfer functions from a small set of measurements [J]. Journal of the Acoustical Society of America, 2012, 132(1): 282-294.
[10] Breebaart J. Effect of perceptually irrelevant variance in head-related transfer functions on principal component analysis [J]. Journal of the Acoustical Society of America, 2013, 133(1): EL1-EL6.
[11] 胡红梅, 周琳, 马浩, 等. 基于偏最小二乘回归的与头相关传递函数的个人化 [J]. 电子与信息学报, 2008, 30(1): 154-158.
[12] Algazi V, Duda R, Thompson D, et al. The CIPIC HRTF database [C]// IEEE ASSP WAS- PAA. 2001: 99-102.
[13] Wu Z, Wang W. Spatial hearing model based on spatial feature extraction and artificial neural network [J]. Chinese Journal of Acoustics, 2000, 19(2): 123-127.
[14] Wu Z, Chan H, Lam F K, et al. A time domain binaural model based on spatial feature extraction for the head-related transfer function [J]. Journal of the Acoustical Society of America, 1997, 102(4): 2211-2218.
[15] He X, Cai D, Partha N. Laplacian score for feature selection [C]// Proceedings of Advances in Neural Information Processing Systems. 2005: 507-514.
[16] Zhang M, Kennedy R, Abhayapala T, et al. Statistical method to identify key anthropo- metric parameters in HRTF individualization [C]// 2011 Joint Workshop on Hands-free Speech Communication and Microphone Arrays (HSCMA). 2011: 213-218.
[17] Nianjan M, Fallsise F. Neural networks and radial basis functions in classifying static speech patterns [J]. Computer Speech and Language, 1990, 4(3): 275-289.
[18] Han L, Wang Q, Zheng F, et al. Improved k-means initial clustering center selection algorithm [J]. Computer Engineering and Applications, 2010, 46(17): 150-152.
[19] 李倩茹, 王于丁, 张晓芳. 一种变遗忘因子 RLS 算法的分析与仿真 [J]. 现代电子技术, 2008, 31(17): 45-47.
[20] 王惠文. 偏最小二乘回归方法及其应用 [M]. 北京: 国防工业出版社, 1999: 150-171.
[21] Musavi M, Ahmed W, Chan K, et al. On the training of radial basis function classifiers [J]. Neural Networks, 1992, 5(4): 595-603.
