基于表面肌电信号的手写识别系统设计

基于表面肌电信号的手写识别系统设计(任务书,开题报告,论文说明书12000字,代码)
摘要
人体表面肌电信号（Surface Electromyographic,sEMG）是由肌肉纤维伸缩运动而产生的一种电位信号，其具有微弱性和复杂性的特性。通过人体表面电极采集sEMG，其作为一种信号源被广泛应用于临床诊断、康复医学，是近年来的研究热点。在sEMG信号的研究过程中，难点在于如何降低信号获取过程中的伴随噪声，如何从信号中提取出有效、准确的特征信息并完成相应手写数字0-9的识别。
本文的研究内容为设计及实现以sEMG为信号源和控制源的手写识别系统。研究的主要内容包括sEMG的采集及预处理、特征提取、模式识别，在此研究基础上，完成基于sEMG的手写识别系统，实现简单的手写数字0-9的识别，主要完成以下几项工作：
1）在信号采集及消噪方面，使用实验室已有设备对信号进行采集，并使用固定阈值小波去噪的方法对信号进行消噪处理。
2）在特征提取方面，选用小波高频系数最大绝对值作为sEMG的特征值，使得特征矢量差异明显，且与信号一一对应。
3）在模式识别方面，对比常用的模式识别分类器，最终选择BP神经网络作为本文的分类器，并对其各参数进行选择以提高其识别率，如网络层数、传递函数、目标误差等。
关键词：sEMG；去噪；特征提取；模式识别

Abstract
Surface Electromyography (sEMG) is a kind of potential signal generated by the telescopic movement of muscle fibers, which has the characteristics of weakness and complexity. It is widely used in clinical diagnosis and rehabilitation medicine as a kind of signal source. It is also a hotspot in recent years. In the process of sEMGsignal research, the difficulty lies in how to reduce the accompanying noise in the signal acquisition process, how to extract the effective and accurate characteristic information from the signal and complete the recognition of the corresponding action 0-9.
The research content of this paper is to design and realize handwriting recognition system using sEMG as signal source and control source. The main contents of the study include sEMG acquisition and pretreatment, feature extraction and pattern recognition. On the basis of this research, we complete the handwriting recognition system based on sEMG and realize the simple manpower gesture action recognition.The paper mainly complete the following work:
1) On the signal acquisition and denoising, the signal is collected by the equipment of school.And select the fixed threshold wavelet denoising method to de-nose the noisy.
2)On the aspect of feature extraction, the maximum absolute value of the wavelet high frequency coefficient is chosen as the eigenvalue of sEMG, which makes the difference of the eigenvalues obvious and corresponds to the signal one by one.
3) On the aspect of pattern recognition, we review the commonly used pattern recognition classifier, and finally choose BP neural network as the classifier of this paper, and design its parameters, such as network layer number, transfer function, target error and so.
Key Words:sEMG;Signaldenoising;Characteristic acquisition ;Pattern recognition
 

目录
第1章绪论    4
1.1研究目的及意义    4
1.2国内外研究现状    5
1.2.1 sEMG采集和预处理现状    5
1.2.2 sEMG特征提取研究现状    6
1.2.3分类器研究现状    7
1.3论文内容安排    7
第2章原理    9
2.1 sEMG的去噪预处理    9
2.1.1 sEMG简介及噪声来源    9
2.1.2小波去噪    9
2.2特征提取    11
2.3 BP神经网络    13
2.3.1 BP神经网络简介    13
2.3.2 BP神经网络训练算法    14
2.3.3 BP神经网络网络参数选择    16
2.4小结    17
第3章实验过程及结果分析    18
3.1 sEMG采集    18
3.1.1实验动作及肌肉选取    18
3.1.2 sEMG的采集设备    18
3.1.3 sEMG的采集过程    19
3.2 sEMG去噪处理    20
3.3 sEMG特征提取    21
3.4 sEMG模式识别    23
3.5本章小结    25
第4章总结与展望    26
4.1总结    26
4.2展望    26
第5章参考文献    27
致谢    28