Article
Assessment of Head-Related Transfer Function Time-Alignment Preprocessing Through Spatial Principal Component Analysis (en)
* Presenting author
Abstract:
Due to the symmetrical shape of the head and ears, the Head Related Transfer Functions (HRTFs) share features across different spatial directions. The HRTF signals are shifted in time and filtered due to the direction of arrival of the incoming sound to the eardrums. To efficiently encode the HRTFs and to overcome the spatial redundancy in high-resolution datasets, preprocessing methods have been previously suggested. Although these methods help to encode the HRTFs, such preprocessing methods are reported to be either sensitive to noise or prone to insert Interaural-Time Difference (ITD) errors into the signal. In this study, a time-alignment preprocessing method based on ITD approximation methods is proposed and evaluated using spatial Principal Component Analysis (sPCA). The HRIR signals are time-aligned by eliminating their relative time delays. Thereafter, the preprocessed HRIRs are encoded using sPCA, and the method efficiency is evaluated by comparing the required principal components to span a certain variation level and reconstruction accuracy of the HRIR signals before and after the preprocessing. The outcomes reveal that the dimension reduction level acquired through sPCA can be considered as an effective indicator to evaluate the performance of the HRTF preprocessing method in spatial encoding applications.