三维声双耳渲染算法的客观评测与分析

Objective Evaluation and Analysis of 3D Sound Binaural Rendering Algorithm

本文对5种双耳渲染算法进行了客观评测。首先对双耳信号提取双耳特征和单耳特征,并进行特征选择和降维,最终选择岭回归建立了总体评价和其二级维度,即音质、定位和空间感的客观评测模型,并探究了主观感知和客观参数之间的关系。实验结果表明,将实验信号时频域特征的双耳差值用于客观评测模型可以获得理想的拟合效果;对于音质的感知,与频域能量相关的特征是最为重要的影响因素;变化丰富的双耳线索和双耳间频谱信息的变化差异性对于定位和空间的感知有较为显著的影响。

This paper conducts an objective evaluation of the rendering effects of five binaural rendering algorithms. Firstly, binaural features and monaural features from binaural signals are extracted and selected based on feature selection, dimensionality reduction. Finally, ridge regression is chosen to build the objective evaluation models of preference and its secondary dimensions, such as timbre, location and spatial impression. Furthermore, the relationship between subjective perception and objective parameters is explored. The experimental results show that using the binaural difference of the time-frequency domain characteristics of the experimental signals for the objective measurement model can obtain an ideal fitting effect;For the perception of timbre, the acoustical features related to the spectral energy are the most important factors;The variable binaural cues and the diverse changes between binaural spectrum information have significant effects on the perception of location and spatial impression. In the future, the measurement and improvement of three-dimensional sound binaural playback should consider the binaural interaction of the time-frequency domain information.