Anxiety State Prediction Of Individual Anxiety Patient Based On Finger Pulse Signal

The outbreak of the COVID-19 pandemic has brought unprecedented challenges to the world.People are generally feeling anxious and uneasy when facing the spread of the virus and potential health risks.Factors such as uncertainty,fear,and social distance restrictions caused by the pandemic can all contribute to triggering anxiety disorders in individuals.In recent years,the increasing incidence of psychological disorders has attracted growing attention from society.According to a report from the World Health Organization,the prevalence of mental illnesses,such as anxiety,has increased by 25%in the first year of the pandemic,and the number of new patients with anxiety disorders worldwide has exceeded 90 million.This kind of anxiety can significantly reduce people’s quality of life and work efficiency.To improve people’s quality of life and work efficiency,promote social stability and vitality,it is particularly important to prevent and treat anxiety disorders and related symptoms in a timely manner.To achieve timely intervention and treatment of anxiety symptoms,it is necessary to accurately assess the patient’s anxiety state quickly.However,the mainstream anxiety assessment methods currently still rely on the clinical experience of doctors and self-rating scales.The self-rating scale methods have the disadvantages of high subjectivity and poor repeatability and cannot be used for real-time monitoring and analysis of the patient’s condition.Therefore,there is an urgent need for an anxiety prediction method based on objective data that can better monitor and manage individual anxiety emotions,and provide personalized anxiety symptom intervention and treatment plans for patients.With the rise of the interdisciplinary field of computational psychiatry,data-driven techniques,including machine learning and deep learning,are broadly adopted to analyze neurobiological pathology.These techniques have immense practical value in the prevention,diagnosis,and intervention of psychiatric disorders.Apart from that,pulse information from the human body can reflect crucial physiological characteristics,which is proved by the treatment of Chinese medicine for thousands of years.Modern collectors of pulse signal are portable to meet the needs of quick measurements.Therefore,it is particularly important to design an algorithm that can predict anxiety states based on deep learning and pulse signals.The algorithm can not only provide support for the early prevention and auxiliary treatment,but also hold enormous promise in the field of intelligent detection of traditional Chinese Medicine.In this paper,we thoroughly investigate the existing methods and propose a rational method for collecting and extracting features of finger pulse signals from anxiety patients.Furthermore,a deep learning algorithm is designed to predict the anxiety status of patients.The main content and results of this paper are as follows:1.Construction of a finger pulse signal dataset with anxiety level and sleep quality labels:Specifically,the pulse signals of the anxiety patient are continuously detected by the photoelectric volume sensor,collected and labeled according to established rules;the pre-processing operations such as noise reduction,baseline drift removal,pulse period interception,etc are then conducted using signal-processing algorithms;and a data equalization method is used to obtain a dataset containing 312 finger pulse signals labeled with anxiety status and sleep quality.2.Research on anxiety level prediction model based on finger pulse signals and deep learning:Aiming at improving the accuracy of existing models for multi-level emotion recognition,a dual model based on DPN is proposed to evaluate the anxiety status based on the finger pulse data.Specifically,a dual-model joint classification approach is proposed,where key features of different classes are learned by different models;and three convolutional neural networks,ResNet,DenseNet,and DPN,are compared for their effectiveness in extracting signal features.Extensive results indicate that the dual model can effectively alleviate the boundary error between categories compared to the single model.Moreover,the dual-path network structure of DPN is proven to have better feature extraction capability for pulse signals.3.Research on anxiety level prediction model based on multi-task learning and attention mechanism:To overcome the limited expressiveness of existing single-task models,the correlation between anxiety and sleep labels is modeled on the basis of the above model,using sleep labels to bring additional information for anxiety level prediction.Concretely,the dataset is validated for label correlation using statistical probabilistic approaches;and the model structure and joint training method are designed to adaptively fuse features using multi-task learning and convolutional block attention module.Empirical experiments show that the auxiliary task can further improve the precision of anxiety evaluation models.