Social Activity Detection


Various sensing systems have been exploited to monitor social activities, which are one of the most important indicators of physical and mental health. However, existing solutions either need to deploy significant ambient infrastructure or cannot provide enough detailed insight regardig the quality of social interactions which is of great interest to psychiatrists and caregivers. In this reserach, we use Body Sensor Networks (BSNs) to accurately detect various social activities and provides insights absent in existing systems.



We present a grammar-based social activity detection framework using cooperative BSNs and speech recognition. In our framework, we detect social activities by trying to answer five questions: What is the subject doing? How many others are close to the subject? What are others in the vicinity of the subject doing? Is there a conversation? What is the sentiment of the conversation? Cooperative BSNs and speech recognition are used to answer these questions. The main contributions of our work are:
1. We construct a framework based on a context-free grammar to define social activities. This imbues our system with the flexibility to handle new types of social interactions effortlessly.
2. Our system not only collects data from an individual BSN as in most existing work, but also employs information from nearby BSNs.
3. Speech recognition is used in our solution, so that not only the presence of conversation is detected, but the sentiment of the conversation is also identified.

Social Activity Detection Process: