Abstract：
In order to improve the prediction accuracy of the sound quality of vehicle interior noise, a novel sound quality prediction model was proposed based on the physiological response predicted metrics, i.e., loudness, sharpness, and roughness. First, a human-ear sound transmission model was constructed by combining the outer and middle ear finite element model with the cochlear transmission line model. This model converted external input noise into cochlear basilar membrane response. Second, the physiological perception models of loudness, sharpness, and roughness were constructed by transforming the basilar membrane response into sound perception related to neuronal firing. Finally, taking the calculated loudness, sharpness, and roughness of the physiological model and the subjective evaluation values of vehicle interior noise as the parameters, a sound quality prediction model was constructed by TabNet model. The results demonstrate that the loudness, sharpness, and roughness computed by the human-ear physiological model exhibit a stronger correlation with the subjective evaluation of sound quality annoyance compared to traditional psychoacoustic parameters. Furthermore, the average error percentage of sound quality prediction based on the physiological model is only 3.81%, which is lower than that based on traditional psychoacoustic parameters.
摘要：
为了提高车内噪声声品质的预测精度,提出了一种基于生理反应预测指标的音质预测模型，即，响度,清晰度,和粗糙度。首先,通过将外耳和中耳有限元模型与耳蜗传输线模型相结合，构建了人耳声音传输模型。该模型将外部输入噪声转换为耳蜗基底膜响应。第二,响度的生理感知模型，清晰度,通过将基底膜反应转化为与神经元放电相关的声音感知来构建粗糙度。最后,采用计算出的响度，清晰度,以生理模型的粗糙度和车内噪声的主观评价值为参数,利用TabNet模型构建了音质预测模型。结果表明，响度，清晰度,与传统的心理声学参数相比，通过人耳生理模型计算的粗糙度与声音质量烦恼的主观评估具有更强的相关性。此外，基于生理模型的音质预测平均误差百分比仅为3.81%,低于传统心理声学参数。
