基于特征选择算法的ECG信号分类

ECG signal classification based on feature selection algorithm

为了提高不同类别心电图(Electrocardiogram,ECG)信号的识别精度,使用小波分析提取心电信号特征,并使用分段距离的特征筛选方法对特征进行筛选排序,去除冗余和干扰特征,挑选出关键特征。通过缩减特征数量,提高分类的精度和效率。结合机器学习分类器对特征进行分类,比较分类效果。结果显示,在MIT-BIH数据集上,本方法的分类精度比不使用特征选择分类精度高0.22%,分类精度最高达到99.67%。试验证明本研究提出的模型能够区分4种常见的ECG信号,较传统方法优势明显。

In order to improve the recognition accuracy of different types of electrocardiogram(ECG) signals,wavelet analysis was used to extract ECG signal features,and the feature screening method of segment distance was used to sort the features,remove redundant and interference features,and select the key features.By reducing the number of features,the accuracy and efficiency of classification was improved.The features were classified by machine learning classifier,and the classification effect was compared.The results showed that on MIT-BIH data set,the classification accuracy of this research method was 0.22% higher than that without feature selection,and the highest classification accuracy was 99.67%.Experiments showed that the model proposed in this study could distinguish four common ECG signals,which had obvious advantages over traditional methods.