Binary sensing and perception for human behavior study

This paper presents a binary sensing and perception framework for human behavior study. The target behavioral information usually includes two parts: (1) average geometry size and (2) invariant motion patterns. Binary sensing is the simplest data acquisition form, which results in low data throughput, communication overhead and energy consumption. However, binary sensing also suffers from several disadvantages: (1) loss of information, (2) nonlinearity of logic operations, and (3) difficulty in pattern analysis. The proposed framework contains following features: (1) bio-inspired sampling geometry, (2) integral geometry based behavioral information acquisition and estimation, and (3) differential geometry based behavior analysis. A number of sensing modalities, such as laser and thermal, have been utilized to demonstrate the performance of the proposed paradigm in human behavior sensing and analysis.