基于加权弹性网络回归的个性化HRTF方法研究  

Research on personalized modeling of head-related transfer function based on weighted elastic network regression

作　　者：冯莹依 刘海生[1] FENG Yingyi;LIU Haisheng(Institute of Acoustics,Tongji University,Shanghai 200092,China)

机构地区：[1]同济大学声学研究所,上海200092

出　　处：《声学技术》2023年第6期832-838,共7页Technical Acoustics

摘　　要：个性化的头相关传输函数(Head Related Transfer Function,HRTF)对于虚拟听觉技术的实现至关重要。然而在具体的应用过程中,测量每一位受试者的个性化HRTF较为繁琐,为此文章提出一种基于加权弹性网络回归的算法,只需获取受试者的生理参数即可获得HRTF的个性化幅度响应。首先通过数据库中的受试者数据,根据生理参数与幅度的相关性计算获得生理参数的权值,并将权值加入到同时含有1范数和2范数的弹性网络回归中,以此来获取新受试者的生理参数稀疏系数;最后将所得稀疏系数与数据库中的HRTF幅度结合就可以得到新受试者的个性化幅度响应。结果表明,文中方法对于个性化HRTF幅度的合成有较好的效果,尤其是在中低频段内准确度较高。Personalized head related transfer function(HRTF)is essential for the realization of virtual auditory.However,in the application process,it is tedious to measure the personalized HRTF for every user,so this paper proposes an easy method based on weighted elastic network regression,which can obtain the personalized amplitude response of HRTF as long as the anthropometric features of the subject are obtained.Firstly,through the database of HRTFs,the weights of the anthropometric features are calculated according to the correlation between the parameters and the amplitude,and the weights are added to the elastic network regression with both 1-norm and 2-norm so as to obtain the sparse coefficient of anthropometric features of the new subject.Finally,the personalized amplitude response of the new subject can be obtained by combining the obtained sparse coefficient with the HRTF amplitude in the database.The experiments show that the method in this paper has a good effect on the synthesis of personalized amplitude response of HRTF,especially in the low and medium frequency bands.

关 键 词：个性化头相关传输函数 生理参数 加权弹性网络回归 

分 类 号：TN912[电子电信—通信与信息系统]