Development of an AI-based Mobile App for Automatic Depression Screening Using Speech in English and Chinese

Current AI-based depression assessment apps provide significant advantages for users and clinicians through personalized, accessible mental health support. However, most rely on text or questionnaire-based methods, which are often influenced by user subjectivity, while speech-based detection remains overlooked. In this study, we developed a deep learning based mobile app, MoodEcho, for screening depression using users' audio recordings. MoodEcho is an iOS-based mobile app developed using cloud technology, consisting of three key modules: an audio recording module, a speech-based depression detection module, and a user management module for monitoring and managing detection results. We extracted Mel-spectrogram features from an English depression speech dataset, DAIC-WOZ, to train a convolutional neural network (CNN) incorporating Squeeze-and-Excitation (SE) modules, achieving an F1-score of 0.86 on the test set. Besides, we tested a Chinese depression speech dataset collected from clinical interviews using the same SE-CNN model, and the F1-score of the test set was 0.75. MoodEcho implements a dual-language detection capability through a language-adaptive modeling framework, demonstrating its effectiveness as a cross-lingual screening tool. These results show MoodEcho's promising potential as a convenient and reliable tool for clinical decision making, providing valuable support for both users and clinicians.Clinical Relevance- This study presents an AI-based depression detection mobile app as an objective, non-invasive, and accessible tool for early screening, periodic screening, continuous monitoring, and personalized management of depression, ultimately supporting clinical diagnostic and improving diagnostic objectivity.