Acoustic Sensing for Multi-User Heartbeat Monitoring Using Dualforming

Acoustic sensing for heartbeat monitoring has emerged as a prevailing research topic in wireless sensing. However, existing acoustic sensing systems face two limitations: a restricted sensing range and operation limited to a single user, impeding large-scale deployment of its applications. In this paper, we present DF-Sense, a <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">D</i>ual <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">F</i>orming based multi-user acoustic <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Sensing</i> system for heartbeat monitoring in home settings. Specifically, we design a novel sensing signal-to-noise ratio (SSNR) enhancement model, namely <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Dualforming</i>, which leverages constructive superposition across multiple subcarriers and microphones. To facilitate <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">Dualforming</i>, we propose a novel MUltiple Subtle SIgnal Classification (MUS<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>IC) method and a 2-D peak identification scheme to locate and identify multiple subjects with subtle motions. Additionally, we propose a phase change-based method to promptly identify body leaning and adaptively re-localize subjects, thereby avoiding the high computational cost. Finally, we propose an enhanced recursive least squares (RLS) filter to effectively reconstruct high-quality heartbeat waveforms from Channel Frequency Response (CFR) signals affected by limb movements. Experimental results show that DF-Sense achieves high precision measurement of instantaneous heart rates within a range of 10 m, sufficient for most daily space requirements, and can monitor heartbeat for up to 6 subjects in a 2-D space.