基于人体测量参数的个性化HRTF建模  

Personalized HRTF Modeling Based on Anthropometric Parameters

作　　者：魏永健 周静雷[1] WEI Yongjian;ZHOU Jinglei(College of Electronic Information,Xi′an Polytechnic University,Xi′an 710600,China)

机构地区：[1]西安工程大学电子信息学院,陕西西安710600

出　　处：《电工技术》2024年第18期118-122,共5页Electric Engineering

摘　　要：头相关传输函数(Head Related Transfer Function, HRTF)在空间音频渲染中具有关键作用,能够显著提升个体的听觉体验。然而,要实现最佳的听觉效果,HRTF必须与受试者的解剖特征相符。为了达到这一目标,采用了一种基于人体测量特征的方法来估算个体化HRTF,特别地,扩展了耳廓的测量参数,并考虑了常被忽略的耳廓腔室对HRTF的影响。通过主动形状模型(Active Shape Models, ASM),自动从特定耳廓的标记点提取耳廓测量参数。接着,使用轻量级梯度提升机(Light Gradient Boosting Machine, LightGBM)模型,根据提取的耳廓测量参数及头部测量参数,预测个体在中垂面上的HRTF幅度。评估结果显示,所提取的耳廓特征能够显著提升HRTF个性化的客观性能指标。The head-related transfer function(HRTF)plays a crucial role in spatial audio rendering,significantly enhancing an individual′s auditory experience.However,for optimal auditory outcomes,the HRTF must align with the subject′s anatomical features.To achieve this,our study employed an anthropometric measurement-based method to estimate individualized HRTFs.Specifically we expanded the measurement parameters of the pinna and considered the effects of the often-overlooked concha cavity on the HRTF.Using active shape models(ASM),we automatically extracted pinna measurement parameters from marked points on specific pinnas.Subsequently,employing a light gradient boosting machine(LightGBM)model,we predicted individual HRTF magnitudes in the median plane based on the extracted pinna and head measurement parameters.The evaluation results indicated that the extracted pinna features significantly improved the objective performance metrics of HRTF personalization.

关 键 词：HRTF个性化 人体测量参数 ASM LightGBM 

分 类 号：TN912.16[电子电信—通信与信息系统]