Infer Daily Mood using Mobile Phone Sensing

With the increasing stress and unhealthy lifestyles in people's daily life, mental health problems are becoming a global concern. In particular, mood related mental health problems, such as mood disorders, depressions, and elation, are seriously impacting people's quality of life. However, due to the complexity and unstableness of personal mood, assessing and analyzing daily mood are both difficult and inconvenient, presenting a major challenge in mental health care. In this paper, we propose a novel framework called MoodMiner for assessing and analyzing mood in daily life. MoodMiner uses mobile phone data - mobile phone sensor data and communication data (including acceleration, light, ambient sound, location, call log, etc.) to extract human behavior pattern and assess daily mood. A factor-graph based model called SFFG is proposed, combining temporal and social information together with mobile phone sensing data for daily mood assessment. Our approach overcomes the problems of subjectivity and inconsistency of traditional mood assessment methods, and achieves a fairly good accuracy (around 70%) with minimal user intervention. We have built a system with clients on Android platform and an assessment model based on factor graph. We have also carried out experiments to evaluate our design in effectiveness and efficiency.