Recognition Of Tongue Diagnosis Constitution Based On Wavelet Attention And Reshape Fusion

Traditional Chinese medicine is used to prevent,diagnose and treat diseases,and it plays a unique role in protecting and enhancing human physical health.Tongue diagnosis is a crucial method in the diagnosis process of traditional Chinese medicine,which can help doctors understand the health status of patients for further diagnosis and treatment.As an extremely important application of tongue diagnosis,constitution recognition has attracted the attention of many scholars.It can identify a person’s constitution by observing tongue indicators such as tongue color,shape,coating color,and humidity,which is helpful for grasping Based on the patient’s health status and individual differences,a personalized treatment plan is developed.In recent years,deep learning has made breakthroughs in many fields,especially in the field of computer vision represented by convolutional neural network technology.At present,scholars have used convolutional neural networks to complete the task of physical fitness recognition on tongue image data and achieved certain results,but these methods still have various shortcomings.First,the representation capability of the convolutional neural network used to extract tongue image features is limited,and the capability of feature extraction needs to be further improved.Second,existing tongue diagnosis methods lack the reuse of multiscale and multi-level features in the network.Finally,in the current research on tongue diagnosis methods,an effective feature fusion method is lacking.Aiming at the shortcomings of the current tongue diagnosis methods and in order to improve the performance of automatic constitution recognition,this paper(1)proposes a Wavelet Attention module,which can obtain multi-scale features through two-dimensional discrete wavelet transform separation.And use the attention mechanism to weighted and fuse the features of multiple scales,which improves the feature extraction ability of the neural network;(2)proposes a Reshape Fusion module for different levels of features,which can mine the associations of multiple levels of features through reshaping operations,and then efficiently fuse the features,so as to more effectively develop and use the rich features in the neural network;(3)A network framework for tongue diagnosis constitution recognition based on Wavelet Attention and Reshape Fusion is proposed.It integrates the two modules of Wavelet Attention and Reshape Fusion into the convolutional neural network to generate more accurate tongue image attributes.Finally,based on this attribute,the task of constitution recognition is completed,and the interpretability of constitution recognition results is increased.Promotion and application of automated tongue diagnosis technology.In this paper,a variety of experiments are carried out on the tongue diagnosis and constitution dataset to verify the effectiveness of the proposed method.Ablation experiments verified the respective effectiveness of the Wavelet Attention module and the Reshape Fusion module,and the comparative experiments verified the effectiveness and efficiency of the constitution recognition framework based on Wavelet Attention and Reshape Fusion.The visualization experiments also showed that this method has certain interpretability.In general,the experimental results show that the methods proposed in this paper can effectively improve the performance of the constitution recognition task,and finally achieve an accuracy of53.95%,which is 3.18% higher than the benchmark model.