车内稳态噪声干扰度有源增益系数提前梯度优化方法
作者:
作者单位:

同济大学,同济大学,同济大学

作者简介:

通讯作者:

中图分类号:

U461.4

基金项目:


An Advance Gradient Optimization Method to Optimize Sound Distraction Levels of a Passenger Vehicle’s Interior Stationary Noise Samples with Active Noise Equalization
Author:
Affiliation:

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    为使用有源噪声均衡技术快速优化车内稳态噪声干扰度,分析了传统枚举方法搜索有源均衡最优增益系数向量用以优化噪声品质的特点;通过主观评价建立了车内20500Hz频率范围内稳态噪声对于人员注意力的干扰程度的噪声品质,干扰度;建立了以20500Hz频率范围内各个临界频带线性总声压幅值为输入的噪声干扰度反向传播(BP)神经网络客观计算模型;推导了以收敛后的BP神经网络权值表示的各个输入对于噪声干扰度的灵敏度和贡献量;推导了一个频带的有源增益系数、有源均衡前幅值频谱、参考信号,与有源均衡后频带内线性总声压幅值的关系;基于这个关系和噪声干扰度的灵敏度以及贡献量提出了搜索最优增益系数的提前梯度优化方法.使用提前梯度方法有源优化车内稳态噪声干扰度,优化过程耗时较少,主观评价试验显示,优化效果较准确,车内稳态噪声干扰度改善较显著.

    Abstract:

    In order to quickly optimize a passenger vehicle’s interior stationary noise samples’ sound distraction levels with active noise equalization, the traditional enumeration method used to search for optimal gain coefficient vector of active noise equalization system and to optimize sound quality was analyzed; subjective evaluation was used to evaluate the distraction levels of the passenger vehicle’s interior stationary noise samples within 20500 Hz; Back Propagation (BP) neural network using the barks’ total sound pressure linear amplitudes within 20500 Hz as inputs was used to fit the noise samples’ sound distraction levels; the trained BP neural network’s weights were used to deduce the network’s inputs’ sensitivities and contributions to the sound distraction levels; an equation was deduced to predict the total sound pressure linear amplitude of a bark after active noise equalization with a given gain coefficient, the original sound pressure amplitude spectrum and the reference signal of the active equalization system; based on this equation, the sound distraction levels’ sensitivities and contributions, an advance gradient optimization method was designed to search for optimal gain coefficient vector and to optimize sound distraction levels of the noise samples. The time consumption of the optimization process is low. Active noise equalization using the gain coefficient vectors acquired by the advance gradient optimization method was executed and the equalized noise samples’ sound distraction levels were evaluated with subjective evaluation. The result shows good accuracy and the sound distraction levels are improved significantly.

    参考文献
    相似文献
    引证文献
引用本文

徐海卿,周鋐,靳畅.车内稳态噪声干扰度有源增益系数提前梯度优化方法[J].同济大学学报(自然科学版),2016,44(3):0427~0433

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2015-03-10
  • 最后修改日期:2016-01-11
  • 录用日期:2015-12-18
  • 在线发布日期: 2016-03-24
  • 出版日期: