Predicting Psychological Crisis via Smartphone
Most psychological crisis are preceded by behavioural and physiological changes. In this project, we aim to detect such warning signals earlier, to prevent psychological crisis before they cause harm. The individual’s smartphone allows to unobtrusively and continuously track various potential precursors, e.g., changes in social interaction, motion patterns or vocal variation. To accurately assess the mental health status from these multiple timeseries, we combine methods of anomaly detection, graph embedding and neural networks.