The proverb "Actions speak louder than words” contains plenty of truth:

Knowing what activities someone has been performing in the past will divulge a lot of information on the behavior, the aims, and the habits of the person, often more detailed than what this person would tell you. This talk will present a body-worn sensor system for recognizing physical activity based on initial sensors that is currently used in several trials in sleep study and psychiatry. The core of this research is the integration of novel signal processing and machine learning techniques to handle incoming sensor data streams as efficiently as possible, and to make the wearable system run as long as possible on a single battery charge.