ENTPE, Ecole Centrale de Lyon, CNRS, LTDS, UMR5513, 69518 Vaulx-en-Velin, France.
Aix Marseille Univ, CNRS, Centrale Marseille, LMA, 13013 Marseille, France.
J Acoust Soc Am. 2024 Oct 1;156(4):2287-2298. doi: 10.1121/10.0030463.
Many objective measurements have been proposed to evaluate sound reproduction, but it is often difficult to link measured differences with the differences perceived by listeners. In the literature, the best correlations with perception were obtained for measures involving an auditory model. The present study investigated simpler measurements to highlight the signal processing steps required to make the link with perception. It is based on dissimilarity evaluations from two previous studies: 1 study comparing 12 single loudspeakers using 3 musical excerpts, 1 study comparing 21 headphones using 2 musical excerpts, and both studies highlighting 2 perceptual dimensions associated with the relative strengths of bass and midrange. The objective approach compared several signal analyses computing the dissimilarity between the spectra of the recorded sound reproductions. The results show that a third-octave analysis can accurately describe the overall dissimilarity between the loudspeakers or headphones and the two underlying perceptual dimensions.
已经提出了许多客观测量方法来评估声音再现,但通常很难将测量到的差异与听众感知到的差异联系起来。在文献中,与感知相关性最好的是涉及听觉模型的测量方法。本研究调查了更简单的测量方法,以突出与感知建立联系所需的信号处理步骤。它基于两项先前研究的可听度评估:1 项研究使用 3 个音乐片段比较了 12 个单一扬声器,1 项研究使用 2 个音乐片段比较了 21 个耳机,这两项研究都强调了与低音和中音相对强度相关的两个感知维度。客观方法比较了几种信号分析,计算了录制声音再现之间的频谱差异。结果表明,三分频分析可以准确描述扬声器或耳机之间以及两个潜在感知维度之间的整体差异。