Background: In recent years, the number of people with anxiety disorders has increased worldwide. Methods for identifying anxiety through objective clues are not yet mature, and the reliability and validity of existing modeling methods have not been tested. The objective of this paper is to propose an automatic anxiety assessment model with good reliability and validity.
Methods: This study collected 2D gait videos and Generalized Anxiety Disorder (GAD-7) scale data from 150 participants. We extracted static and dynamic time-domain features and frequency-domain features from the gait videos and used various machine learning approaches to build anxiety assessment models. We evaluated the reliability and validity of the models by comparing the influence of factors such as the frequency-domain feature construction method, training data size, time-frequency features, gender, and odd and even frame data on themodel.
Results: The results show that the number of wavelet decomposition layers has a significant impact on the frequency-domain feature modeling, while the size of the gait training data has little impact on the modeling e ect. In this study, the time-frequency features contributed to the modeling, with the dynamic features contributingmore than the static features.Ourmodel predicts anxiety significantly better in women than inmen (rMale = 0.666, rFemale = 0.763, p < 0.001). The best correlation coe cient between the model prediction scores and scale scores for
all participants is 0.725 (p < 0.001). The correlation coe cient between themodel prediction scores for odd and even frame data is 0.801∼0.883 (p < 0.001). 

Conclusion: This study shows that anxiety assessment based on 2D gait
video modeling is reliable and e ective. Moreover, we provide a basis for the development of a real-time, convenient and non-invasive automatic anxiety assessment method.