论文总字数：18666字

摘 要

随着汽车行业不断地发展，人们对出行时，汽车乘坐的要求愈发提高，轿车内噪声作为舒适度中最重要的一项指标，噪声严重影响着人们的生活和健康，也越来越引起人们的重视。发动机工作时所引起的车身振动，以及车身板壁投射的排气噪声和机械噪声已经成为车内低频噪声的主要因素。随着轿车保有量的日趋增长，汽车运行时产生的噪声问题愈发严重，国内外都制定了相应的法律法规用于限制汽车的排气噪声。随着深度学习技术的迅速发展，将深度学习技术运用到消除排气噪声上有着积极的意义。本课题着重研究将深度学习运用到噪声控制系统上，并对某款汽车噪声进行抑制，尽可能消除汽车排气时的噪声。

本课题以树莓派4为核心控制器，以实现噪声的抑制效果，该系统由树莓派、免驱USB麦克风及Python3中NumPy，Matplotlib，Neurolab包等构成。它首先通过麦克风获取现场的噪声，然后通过树莓派处理后得出BP神经网络参数，最后对后续环境噪声进行处理。

关键词：神经网络；反向传播算法；深度学习；噪声抑制；树莓派

Car interior noise suppression based on deep learning technology

ABSTRACT

With the continuous development of the automobile industry, people’s requirements for comfort when riding in a car are increasing. The noise in a car is the most important indicator of comfort. Noise seriously affects people’s lives and health, and it also causes more and more problems. People's attention. The vehicle body vibration caused by the engine working, as well as the exhaust noise and mechanical noise projected by the body panels have become the main factors of low-frequency noise in the car. With the increasing number of cars, the noise problem generated during car operation has become more and more serious. Corresponding laws and regulations have been formulated at home and abroad to limit the exhaust noise of cars. With the rapid development of deep learning technology, it is of positive significance to apply deep learning technology to eliminate exhaust noise. This topic focuses on the application of deep learning to the noise control system, and suppresses the noise of a certain type of car, and eliminates the noise when the car is exhausted as much as possible.

This topic uses Raspberry Pi 4 as the core controller to achieve noise suppression. The system is composed of Raspberry Pi, drive-free USB microphone and Python3 NumPy, Matplotlib, Neurolab package, etc. It first obtains the noise of the scene through the microphone, and then obtains the BP neural network parameters after processing by the Raspberry Pi, and finally processes the subsequent environmental noise.

Keywords：neural network, back propagation algorithm, deep learning, noise suppression, Raspberry

目录

摘要 I

ABSTRACT II

第一章 引言 1

1.1设计背景及意义 1

1.2 国内外研究现状 1

1.3 论文主要内容 2

第二章 自适应噪声主动控制的原理 3

2.1 噪声抵消的主要原理 3

2.2 自适应滤波器原理 3

2.2.1 自适应滤波器 3

2.2.2 自适应滤波器模型分析 4

2.3 本章小结 5

第三章 用于噪声抵消的自适应算法的比较 6

3.1 LMS算法 6

3.1.1 LMS算法的原理 6

3.1.2 LMS的优缺点 6

3.2 BP算法 6

3.2.1 BP算法的基本原理 6

3.2.2 神经网络结构、前向传播、反向传播 7

3.3 本章小结 9

第四章 LMS算法与BP神经网络性能的比较 10

4.1 基于LMS算法的自适应噪声消除仿真 10

4.2 基于BP神经网络的噪声消除仿真 11

第五章 在树莓派上部署基于BP神经网络的消噪系统 13

5.1 树莓派4B开发板 13

5.1.1 主要性能 13

5.1.2 树莓派的结构 13

5.2 Python3及依赖包的简介 14

5.2.1 NumPy 14

5.2.2 matplotlib 14

5.2.3 neurolab 14

5.3 编写Python代码部署树莓派 14

5.3.1 使用Python3编写的BP神经网络的调试 15

第六章　总结与展望 16

致谢 17

参考文献 18

附录 19

MATLAB仿真代码 19

LMS算法函数代码 19

LMS算法滤波代码 19

剩余内容已隐藏，请支付后下载全文，论文总字数：18666字