Device independent activity monitoring using smart handhelds

Sensors embedded in smartphones, tabs can be extremely useful in providing reliable information on people's activities and behaviors, thereby ensuring a safe and sound living environment. Activity monitoring through posture identification is increasingly used for medical, surveillance and entertainment (gaming) applications. Major challenges for this task include making the task device independent, use of minimal number of sensors, position of the device, efficient feature extraction etc. Existing works mostly uses one or m ore specific devices for activity monitoring and does not focus on device independence. Ensuring energy efficiency through inexpensive feature extraction technique is another motivation. Consequently, in this paper, a machine learning based activity monitoring framework is proposed that provides device independence using inexpensive time domain features. Implementation of the framework with real devices indicates 96% accuracy with logistic regression when time domain features are used.