Scholarlink

<p>Psychoacoustic metrics (PMs) are quantitative representations of auditory perception that commonly include loudness, sharpness, roughness, fluctuation strength, and tonality. Unlike traditional physical quantities (such as sound pressure level), they reflect the complex nonlinear relationship between acoustic signals and auditory perception to a greater extent. Algorithms for PMs have been developed and integrated into computational tools, but discrepancies exist in results from different tools, and no systematic evaluation has been conducted to date. Four tools were used to calculate PMs for both standard and measured sound stimuli. The accuracy of results from different tools was verified using standard stimuli, and differences in results for measured samples were analyzed using two evaluation methods: statistical test method and perceptual difference method. The impact of these differences was assessed through sound quality evaluation tasks. The perceptual difference method complements statistical tests. The joint deviation in loudness and modulation metrics (fluctuation strength and roughness) can inflate the annoyance discrepancy to 8.8 points on an 11-point rating scale. End-to-end sound quality models avoid prediction accuracy drops due to PM calculation discrepancies. Discrepancies in PM calculations across tools significantly affect annoyance ratings and prediction accuracy. The perceptual difference method can be used as a complement to statistical analysis, and end-to-end models offer robustness against calculation discrepancies.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Inter-Tool Metrology of Psychoacoustic Metrics: Impact on Post-processing Consistency and Sound Quality Models

  • 期刊论文

  • 发表日期：

    2026年6月18日

查看全文
  • Fancheng Liu,
  • Lina Liu,
  • Ke’an Chen,
  • Bingqing Hou,
  • Hao Li,
  • Jun Zhang,
  • Tong Gao

摘要

Psychoacoustic metrics (PMs) are quantitative representations of auditory perception that commonly include loudness, sharpness, roughness, fluctuation strength, and tonality. Unlike traditional physical quantities (such as sound pressure level), they reflect the complex nonlinear relationship between acoustic signals and auditory perception to a greater extent. Algorithms for PMs have been developed and integrated into computational tools, but discrepancies exist in results from different tools, and no systematic evaluation has been conducted to date. Four tools were used to calculate PMs for both standard and measured sound stimuli. The accuracy of results from different tools was verified using standard stimuli, and differences in results for measured samples were analyzed using two evaluation methods: statistical test method and perceptual difference method. The impact of these differences was assessed through sound quality evaluation tasks. The perceptual difference method complements statistical tests. The joint deviation in loudness and modulation metrics (fluctuation strength and roughness) can inflate the annoyance discrepancy to 8.8 points on an 11-point rating scale. End-to-end sound quality models avoid prediction accuracy drops due to PM calculation discrepancies. Discrepancies in PM calculations across tools significantly affect annoyance ratings and prediction accuracy. The perceptual difference method can be used as a complement to statistical analysis, and end-to-end models offer robustness against calculation discrepancies.