摘要
针对脉搏信号频域分析的精度不够、缺乏对研究对象的稳定性分析、提取频域特征的方法趋于经验化等问题,使用小波多分辨率分解,提取精细尺度上的频域特征,结合Lasso套索回归,挖掘脉搏信号频域中蕴含的疾病特征信息.对一类典型心血管疾病患者的脉搏信号进行分析,验证了频谱特征在时域上的稳定性.以房颤和冠心病为例,提取出疾病特征频带并对其进行分类,据此建立Lasso线性分类模型,实现两类疾病的自动识别.
Analysis of pulse signals has the following problems: low accuracy of frequency domain analysis, lack of stability analysis and over-dependence on experience to extract features in frequency domain. Combined with Lasso regression, the wavelet multi-resolution decomposition was used to extract features in precise frequency bands in order to mine the diseases features of pulse signals in frequency domain. A class of pulse signals of patients with cardiovascular diseases was analyzed to verify the stability of frequency features in time domain. The analysis of atrial fibrillation and coronary heart diseases was implemented to show bow to extract feature bands and classify the two diseases. Then a linear classification model was constructed to realize the automatic identification of both diseases.
出处
《浙江大学学报(工学版)》
EI
CAS
CSCD
北大核心
2012年第10期1866-1871,共6页
Journal of Zhejiang University:Engineering Science
基金
国家自然科学基金资助项目(60574079)
浙江省教育厅科研资助项目(Y201017866)
浙江省科技厅科技计划资助项目(2011C23097)
浙江省基金资助项目(LY12F03023)
关键词
脉搏信号
小波多分辨率分解
Lasso回归
特征降维
pulse signal
wavelet multi-resolution decomposition
Lasso regression
dimension reduction offeature