| 许则鹏,刘祖斌,舒勤业,林静,邱丽.面向工业缝纫机噪声品质评估的双耳融合模型[J].声学技术,2024,43(4):511~519 |
| 面向工业缝纫机噪声品质评估的双耳融合模型 |
| The binaural fusion model for the noise quality evaluation of industrial sewing machines |
| 投稿时间：2023-11-15 修订日期：2023-12-27 |
| DOI：10.16300/j.cnki.1000-3630.2024.04.009 |
| 中文关键词: 工业缝纫机 声品质 蒙特卡洛仿真 多元线性回归 双耳融合模型 |
| 英文关键词: industrial sewing machines sound quality Monte Carlo simulation multiple linear regression binaural |
| 基金项目:国家自然科学基金资助项目(61601407)、浙江省自然科学基金资助项目(LY20A040007)。 |
|
| 摘要点击次数: 123 |
| 全文下载次数: 74 |
| 中文摘要: |
| 为解决单纯采用A计权声压级判定工业缝纫机噪声品质好坏的准确率低、采用单次随机划分数据集建模导致模型泛化性不高以及仅使用单通道数据描述声品质导致无法充分反映人的双耳听觉特性的问题，提出一种工业缝纫机噪声品质评估方法。以声压级、A计权声压级、响度、尖锐度、抖动度、粗糙度和音调度作为声品质客观参数，采用基于排序法的参考等级评分法进行主观评价试验，通过蒙特卡洛仿真和多元线性回归建立工业缝纫机噪声品质评估的左右耳独立预测模型和双耳融合预测模型。结果表明：利用蒙特卡洛仿真得到的模型具有最大的统计概率，与单次随机划分数据集进行模型搭建相比，该方法得到的模型具有更小的随机性和更大的泛化性；左右耳独立模型的测试准确率分别为93.56%和92.56%，而双耳融合模型为94.53%，说明双耳融合模型的预测结果与人的主观评价结果更匹配。 |
| 英文摘要: |
| To solve several problems in current noise quality evaluation, such as low accuracy in using only Aweighted sound pressure level to determine the noise quality of industrial sewing machines, low model generalization due to using only single random partitioning dataset for modeling, and failure to adequately reflect the binaural hearing characteristics of human due to using only a single-channel data, a noise quality evaluation method for industrial sewing machines is proposed in terms of objective parameters of sound quality. Firstly, the objective parameters, including sound pressure level, A-weighted sound pressure level, loudness, sharpness, fluctuation strength, roughness and tonality, are calculated. Secondly, the subjective evaluation test is carried out by using the sorting method based reference grade scoring method. Finally, the left-ear and right-ear independent prediction model and the binaural fusion prediction model for the noise quality evaluation of industrial sewing machines are established by Monte Carlo simulation and multiple linear regression. The results show that: the model obtained by using Monte Carlo simulation has the largest statistical probability, and has less randomness and greater generalization compared with the model formed by single random partitioning dataset; the test accuracies of the left and right ear independent models are 93.56% and 92.56%, respectively, while 94.53% for the binaural fusion model, which indicates that the prediction results of the binaural fusion model are better matched with the human subjective evaluation results. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
