摘要
目的研究脉象信号识别模型的建立,充分利用脉象信号在时域中的形态信息,为脉诊客观化研究提供了一种新的思路和方法。方法通过对脉象信号进行预处理和序列规正化,获得长度一致的脉象信号时间序列,利用基于时序卷积的深度学习网络实现对序列形态的特征提取,并建立脉象信号识别模型。结果通过网络自学习提取的特征多数具有显著性差异,7种脉象的平均识别准确率达到85.76%,与常用的脉象识别方法相比有明显提升。结论基于时间序列和时序卷积网络的脉象信号识别方法能够较好地区分不同脉象信号的形态信息,在多种类脉象的识别任务中具有一定的实用价值。
Objective To establish a pulse signal recognition model,which makes full use of the morphological information of the pulse signal in the time domain,provides a new idea and method for the study of pulse taking objectification.Methods We obtained consistent length pulse signal time series by preprocessing and sequence regularization,and used deep learning network based on temporal convolution to realize feature extraction of sequence morphology and establish pulse signal recognition model.Results Most of the features extracted by network selflearning showed significant differences,and the average recognition accuracy of the seven pulse manifestations reached85.76%,which is a significant improvement compared to the commonly used pulse recognition methods.Conclusion The pulse signal recognition method based on time series and temporal convolutional networks can better distinguish the morphological information of different pulse signals,which is of practical value in the task of recognizing multiple types of pulses.
作者
朱光耀
颜建军
郭睿
王忆勤
燕海霞
Zhu Guangyao;Yan Jianjun;Guo Rui;Wang Yiqin;Yan Haixia(School of Mechanical and Power Engineering,East China University of Science and Technology,Shanghai 200237,China;Comprehensive Laboratory of Four Diagnostic Methods,Shanghai University of Traditional Chinese Medicine,Shanghai 201203,China)
出处
《世界科学技术-中医药现代化》
CSCD
北大核心
2021年第9期3056-3064,共9页
Modernization of Traditional Chinese Medicine and Materia Medica-World Science and Technology
基金
国家自然科学基金委员会面上项目(82074332):基于脉象信息集成学习的动脉粥样硬化性心血管疾病发病风险与冠脉危险事件评估模型的研究,负责人:郭睿
国家自然科学基金委员会面上项目(81673880):基于中医四诊大数据的冠心病风险评估与预测模型研究,负责人:王忆勤
上海市科学技术委员会生物医药仪器专项(19441901100):基于人工智能的新型中医脉诊仪的研制,负责人:郭睿
关键词
脉象识别
形态信息
特征提取
时间序列
时序卷积网络
Pulse signal recognition
Morphological information
Feature extraction
Time series
Temporal convolutional network
