In recent years, mental health (especially depression) of university students has aroused general concern. Fast detection of students at risk of depression using multimedia data is a challenge. However, existing methods require the cooperation of participants such as using their speech or facial expression, which are inconvenient to collect and difficult for large-scale screening. In this article, we propose an integrated gait assessment framework that contains the collection and analysis of multimedia data to assess risk of depression for postgraduate students. First, the rigid-body representation is realized by analyzing kinetic energy (KE) and potential energy (PE) generated during walking. Then, we use the fast Fourier transform to analyze KE and PE in the frequency domain for extracting the joint energy feature. Compared with the conventional methods, our method has significantly increased the objectivity of depression assessment in both clinical theory and practice.