Compressive Sensing Based Indoor Human Positioning Using A Single Thermopile Point Detector

Abstract

Compressive sensing becomes more and more popular in recent years as it offers a new paradigm for recovering sparse signals from fewer samples than that needed by the Nyquist sampling theorem. This technology shows powerful potential in image systems, radio frequency, communication and signal processing. Recent works about indoor human detection systems based on compressive sensing raise high attention but are suffered from high cost and large deployment effort. In this paper, a low-cost indoor human positioning system that consists of a single thermopile point detector and a rotating mask is proposed based on compressive sensing. A conventional thermopile point detector could only convert the received infrared radiation within the field of view (FOV) into electrical signals. Our approach is to use a measurement mask to compressively sample the radiation within the area of interest. The mask is designed following the pattern of a binary random matrix. Each row of the matrix forms a submask. For each measurement, one sub-mask encodes the radiation within the FOV in a binary manner. With few times of measurements, a compressed signal sequence is obtained. Then, recovery algorithms are implemented to recover the original spatial distribution of the infrared radiation within the FOV from the compressed signal sequence. Due to the existence of the noise of the thermopile sensor, a denoising recovery method is also developed to improve the recovery quality. From the recovered results, the locations of the human can be determined. The results show that the zonelevel positioning can reach high accuracy while the proposed positioning system remains cost competitive and easy to deploy, which has large potential in applications such as indoor monitoring, healthcare and customized heating and cooling system.