HRIR personalisation using support vector regression in independent feature space
Qinghua Huang, Qiubing Zhuang · Electronics Letters · 2009
A two-stage approach is proposed to obtain an individual head-related impulse response (HRIR). Independent component analysis is first applied to extract independent features and obtain the weight vectors which are then used as the outputs of support vector regression for constructing a personalisation model. The proposed algorithm has achieved better performance for small training samples.