Activity recognition has attracted increasing attention in recent years due to its potential to enable a number of promising context aware applications. Most approaches to activity recognition are based on supervised statistical machine learning methods such as HMM, CRF, and Bayesian networks. It means that we need qualitative and quantitative amount of labeled training data. Collection of annotated activity data still remains one of the main obstacles to activity recognition. In order to resolve mentioned problem, we need a novel way to collect and annotate activity training data from real life. From communities of psychology and cognitive science, Experience sample method(ESM) developed to try to understand people's behavior. They developed various points of views to collect activity data with error such as omission and confusion of order of activity. Our approach is to develop a computerized ESM on mobile devices with simple planner and recognizer to reduce burden of manual input from users. |
|
Research Interestes >
Computerized Experience Sampling Method |