Statistical method to identify key anthropometric parameters in hrtf individualization

This paper identifies the main anthropometric parameters which strongly influence the head-related transfer functions (HRTFs) in a direct physical way using statistical analysis on HRTF measured data. Principle component analysis is separately performed on the head-related impulse responses of all subjects at each direction for each ear to extract the individual information. Then the individual information, along with all anthropometric parameters, is introduced in the multiple linear regression analysis, where F statistic and t statistic are used to characterize the key parameters having strong direct physical effect on the HRTFs. Ultimately, combining with the results of the analysis of the inter-parameter correlations, only eight anthropometric parameters out of 27 are identified as the crucial elements in the role of spatial localization, which provides a guide for efficient HRTF individualization using key anthropometric parameters.