头相关传输函数的数据库匹配方法

Database Matching of Head-related Transfer Functions

针对如何快速获取不同个体的头相关传输函数(Head-related transfer function,HRTF)(即个人化)这一关键问题,本文从生理结构的相似性和相关性出发,提出了一种基于主成分分析、多元线性回归的HRTF数据库匹配方法,并分别将匹配结果和由KEMAR测量得到的数据用于实现双耳可听化,再进行主观听音实验。结果表明,利用数据库匹配得到的HRTF可以有效地提高声源定位精度和降低前后混淆率。

To individualize head-related transfer function（HRTF） fast and effectively,this paper presents a database matching method based on the similarity and the relativity of anthropometric structures,combined with the methods of principal component analysis（PCA） and multiple linear regression（MLR）.The matched HRTF is compared with that of KEMAR dummy head in the subjective listening experiments.Results show that the matched HRTF can enhance the localization accuracy and abate the rate of front-back confusion.