相干声与环境声提取方法的主客观评价关联研究

The Association Study of Objective and Subjective Comparison in Primary-ambient Extraction Methods

相干声与环境声的提取有助于实现灵活的空间声重放。不同方法的提取效果需要通过主观测听评估,但是主观测听耗时长、效率低,不利于实时调整算法。客观评价与主观测听相关联,通过客观指标反映主观评价,有利于优化算法、提高效率并保证算法评估的可靠性。本文对已有的四种典型提取方法(主成分分析法、最小二乘法、掩蔽法以及环境声相位估计法)进行主客观评估,其中对比了不同方法提取成分的提取误差和通道间相关值两个客观指标,并将提取成分用于双耳渲染对音质和声像宽度进行主观测听。主客观评估结果表明,提取成分越精确,在双耳渲染中可得到越好的音质;提取的环境声的通道间去相关性越强,在双耳渲染中声像宽度越宽。

The primary-ambient extraction is helpful to realize flexible spatial sound playback.The effects of different extraction methods need to be verified by subjective evaluation,which is time-consuming,inefficient and not conducive to adjust while operating.If objective comparison is correlated with subjective evaluation,using objective comparison rather than subjective evaluation can improve the efficiency of algorithms and ensure the reliability of the algorithm evaluation.This paper presents the objective comparisons and subjective evaluations on four typical extraction methods,which are Principal Component Analysis(PCA),Least-Squares(LS),Masking and Ambient Phase Estimation with a Sparsity constraint(APES).Extraction performance is quantified by two objective criteria,which are Error-to-Signal Ratio(ESR)and Inter-channel Cross Coherence(ICC).And the extracted components are also used in the binaural rendering to evaluate the sound quality and the sound image width by subjective evaluation.The results show that the extraction methods with less extraction errors can achieve better sound quality in binaural rendering,while the extracted ambient components with weaker inter-channel cross correlation can acquire wider sound images.