VibeAlign: Maximizing <u>V</u> ar <u>i</u> ation of Heart <u>Be</u> at in CSI of COTS Wi-Fi Devices via Signal <u>Align</u> ment

In heartbeat sensing using Commercial Off-The-Shelf (COTS) WiFi devices with common antennas, the channel state information (CSI) for vital sign detection involves an overwhelming static signal (via the line-of-sight path) and a weak dynamic signal (reflected by human body). Conventional methods that compute amplitude or phase from the I/O (in-phase and quadrature) components of CSI involve nonlinear operations, which makes it challenging to detect very weak heartbeat signals without the aid of specialized setups, such as directional antennas. To solve this problem, this paper proposes VibeAlign, a heartbeat sensing system that supports a deployment with ordinary antenna setups, and employs a series of processes to maximize the variation of heartbeat signals. Specifically, to reduce the impact of nonlinear superposition, we directly exploit the I/O components of CSI as a linear combination of dynamic and static components, which enables the extraction of weak heart beat variations. Then, a series of signal alignment processes are proposed to enhance the detection of the heartbeat component, as follows: (i) Ambient noise reduction by aligning CSI in the frequency domain using PCA, (ii) Automatic Gain Control (AGC) impact reduction by aligning the amplitude of CSI, (iii) Maximizing the variation of heartbeat in either I or O using initial phase alignment by rotating I/O components, (iv) Blind spot reduction by avoiding phase misalignment between antennas. Experiment results confirm that the proposed method helps to better detect the weak heartbeat signal, and extends the detection range as well.