Human-Activity Recognition (HAR) Everywhere: Embedded, Mobile, Desktop, Home & Cloud

Human Activity Recognition (HAR) has been a topic of investigation for multiple decades. Researchers have explored a variety of approaches including a service running in the background on a mobile phone, cameras in the environment, RFID readers on people, and many more. In this note, we advocate that for HAR to make the leap from interesting research area to a mainstream technology that enhances our day-to-day experiences with computing it must utilize many parts of a person’s computing ecosystem: from low-power embedded hardware that runs continuously and can be mounted directly in a phone to a service in the Cloud that tracks a person’s activities over the long-term and computes current activity from every available input. We provide a brief discussion of how such a system might be constructed and used.