Abstract:

In the wrist pulse analysis and recognition, non-linear information of wrist pulse signal is difficult to be obtained by time domain or frequency domain analysis methods. Traditional machine learning methods use features extracted manually and cannot realize self-learning of features. A wrist pulse signal analysis and recognition method based on non-threshold recurrence plot and convolutional neural network has been proposed. Based on the nonlinear dynamics theory, the wrist pulse signals are converted into non-threshold recurrence plots. The VGG-16 convolutional neural network is used to extract the nonlinear features of the recurrence plot, and established a wrist pulse classification model automatically. The experimental results show that the classification accuracy of our method is 98.14%. The accuracy is higher than other existing wrist pulse classification methods. The research provides a new method for the classification of wrist pulse signal, and it is practical for wrist pulse diagnosis.

Key words: wrist pulse, non-threshold recurrence plot, convolutional neural network, nonlinear analysis

摘要：

在脉象信号分析识别中，时域、频域等分析方法难以挖掘脉象信号的非线性信息，且传统机器学习方法需要人工定义特征，无法进行特征的自学习。提出一种基于无阈值递归图和卷积神经网络的脉象分析识别方法。基于非线性动力学理论，将脉象信号转换为无阈值递归图，通过VGG-16卷积神经网络实现递归图非线性特征的自动提取，并建立脉象分类模型。实验结果表明，该方法分类准确率可达98.14%，与已有的脉象分类方法相比有所提升。该研究为脉象信号分类提供了一种新的思路和方法，对脉诊客观化具有一定的实用价值。

关键词: 脉象, 无阈值递归图, 卷积神经网络, 非线性分析