Low-order modelling of head related transfer functions based on spectral smoothing and principal component analysis
Michał Pec, Paweł Strumiłło · Proceedings of SPIE, the International Society for Optical Engineering/Proceedings of SPIE · 2009
The article presents wavelet-based spectral smoothing and Principal Component Analysis as a pre-processing step for Head Related Transfer Function (HRTF) filter design. Spectral smoothing of HRTF by means of Stationary Wavelet Transform discards low-value peaks and notches in HRTF which can be treated as noise. The Principal Component Analysis applied to the HRTF data extracts most important frequency components responsible for sound source localization in azimuth and elevation. These methods used in combination are used to reduce the order of filters that are applied in 3D audio rendering system.