Orthogonal-view based compressive motion classification using pyroelectric infrared sensors

This paper presents a compressive classification approach for human motion by using pyroelectric infrared (PIR) sensors. We represent a human motion as a spatio-temporal energy sequence (STES) which is extracted from an infrared radiation domain. The proposed approach consists of two major parts: a compressive sensing unit and an orthogonal-view sensing layout. Through a modulation of the sensor´s field of view, the compressive sensing unit can directly extract the STES. Through the orthogonal-view sensing layout, a fusion of compressive measurements of the STES can provide a more discriminative feature of 3D human motion. A Gaussian Mixture Hidden Markov Model classifier is employed for motion classification in the compressive measurement domain. In this study, PIR sensor arrays and visibility masks are used for compressive sensing. The performance of the proposed approach is evaluated through experiments of upper limb motion classification.