MMDA: A Multimodal Dataset for Depression and Anxiety Detection

In recent years, with the development of artificial intelligence, the use of machine learning and other methods for mental illness detection has become a hot topic. However, obtaining data on mental disorders is very difficult, limiting the development of this field. In this paper, we provide a Multimodal dataset for Depression and Anxiety Detection(MMDA). All subjects in the dataset were diagnosed by professional psychologists, and the subjects’ disorders were determined by combining HAMD and HAMA scores. The dataset includes visual, acoustic, and textual information extracted after de-identification of the original interview videos, for a total of 1025 valid data, which is the largest dataset on mental disorders available. We detail the correlations between each modal and symptoms of depression and anxiety, and validate this dataset by machine learning methods to complete the two tasks of classification of anxiety and depression symptoms and regression of HAMA and HAMD scores. We hope that this dataset will be useful for establishing an automatic detection system for mental disorders and motivate more researchers to engage in mental disorders detection.