WiFi Sensing and Radio-Wave Human Tracking: A Field Overview

Radio waves pass through walls. That simple fact is the foundation of an entire field of research that has grown quietly but substantially over the past decade. Where cameras stop at surfaces and wearables require contact, radio-based sensing keeps working. It tracks people through walls, infers posture, and captures subtle physiological signals, all from signals already present in the environment.

This post surveys the most significant research directions and projects in WiFi sensing and radio-wave human tracking. It is a field worth understanding carefully.

Key Takeaways

• Radio waves penetrate walls and reflect off human bodies, enabling contact-free tracking and vital sign monitoring from standard WiFi hardware.
• Dense pose estimation uses deep learning to map radio reflections onto detailed body surface models, not just skeletal frameworks.
• Passive sensing can detect breathing and heartbeat sub-millimetre chest displacement without wearable devices.
• Dedicated radar systems (UWB, FMCW) offer better performance than WiFi but require additional infrastructure.
• Privacy, generalisation across environments, and deployment reliability remain open challenges in the field.

What Makes Radio-Based Sensing Possible

Radio waves are part of the electromagnetic spectrum. They reflect, scatter, and change phase when they interact with physical objects, including the human body. When a person moves through a space, those movements alter the way radio signals propagate. By analysing those alterations, it becomes possible to infer location, posture, activity, and even physiological state.

The Institute of Electrical and Electronics Engineers (IEEE) has published extensively on channel state information (CSI), which is the per-subcarrier amplitude and phase data available from commercial WiFi hardware. CSI carries far more spatial and temporal detail than simple received signal strength, and extracting it from off-the-shelf routers became one of the catalysts that opened this field to academic researchers without specialised radar equipment.

That is the key point. Much of this work does not require custom hardware. Standard 802.11 WiFi routers, operating at frequencies already licensed for consumer use, can serve as the transmitter and receiver in a passive sensing system.

Dense Pose and Skeletal Tracking Through Walls

One of the more striking branches of this research is through-wall human pose estimation. The goal is to recover a detailed model of the human body's position and orientation from radio reflections alone.

The WiFi-based dense pose direction attempts to map radio reflections onto a detailed surface representation of the human body, recovering not just a stick-figure skeleton but a mesh-like estimate of body surface and posture. This is computationally demanding. It requires deep learning models trained to bridge the gap between radio-frequency observations and the kind of dense body representations that vision-based systems produce.

Neuralase is working in this space, exploring how ambient radio signals and advanced signal processing can be used to infer human body position and movement. The underlying challenge is significant: radio waves do not produce images. They produce interference patterns, and translating those patterns into a meaningful spatial model of a human body requires both careful signal engineering and substantial machine learning work.

Several university research groups have demonstrated that cross-modal learning, training a model on paired camera and radio data, allows the radio branch of the model to eventually operate without the camera at test time. The camera teaches the system what poses look like; the radio learns to predict them. Once trained, only the radio is needed.

Breathing, Heartbeat, and Vital Sign Monitoring

Each breath produces a measurable displacement of the chest wall. That displacement alters the phase of reflected radio signals in a way that sensitive receivers can detect. The effect is small, but modern radio sensing techniques can resolve sub-millimetre motion, which places breathing and even cardiac-induced chest vibrations within reach.

This application sits at the intersection of sensing and health. Passive vital sign monitoring means no wearable device, no chest strap, no patch attached to skin. The room itself does the measuring.

Neuralase is developing technology in this area, using ambient WiFi signals to synthesise physiological signals without contact. The implications for home care, hospital environments, and long-term health monitoring are real. Continuous monitoring has historically required either clinical-grade equipment or consumer wearables. A passive, infrastructure-based approach changes that equation. We have explored WiFi heart rate monitoring to demonstrate the viability of this approach.

Gesture Recognition and Activity Classification

Human tracking does not always mean localisation. Sometimes the goal is classification: is this person walking, sitting, falling, exercising? Radio-based activity recognition has been demonstrated across a wide range of categories.

Gesture recognition is a tighter problem. The movements involved are small and fast, and distinguishing one hand gesture from another requires high temporal resolution and careful feature extraction. Research groups have demonstrated WiFi-based gesture recognition with reasonable accuracy, though performance degrades with distance, multipath complexity, and the presence of multiple people.

Multi-person tracking is genuinely hard. When two or more people are present, their reflections overlap and interfere. Separating individual contributions from the composite signal requires either careful antenna design or sophisticated blind source separation techniques applied in software.

Radar-Based Approaches: Beyond WiFi

WiFi is not the only radio technology used in this space. Dedicated radar systems, including ultra-wideband (UWB) radar and frequency-modulated continuous wave (FMCW) radar, offer performance characteristics that commodity WiFi hardware cannot match. They provide finer range resolution and are designed explicitly for sensing rather than communication.

The Federal Communications Commission (FCC) has allocated spectrum for UWB devices and established technical standards governing their operation, which has helped legitimate commercial and research development in short-range radar sensing. UWB's fine time resolution makes it well-suited to precise localisation and gesture capture.

The tradeoff is cost and accessibility. WiFi-based approaches use hardware already deployed in almost every building. Dedicated radar requires additional infrastructure. For research purposes, WiFi is often the entry point; for production deployment, dedicated radar may offer better reliability.

Where the Field Is Going

The trajectory here is toward systems that understand human presence in space at a level of detail that was not previously achievable without cameras or wearables. Posture, location, activity, breathing, even emotional or cognitive states inferred from physiological proxies. All from radio signals.

Several open questions remain. Privacy is one. A system that can infer what a person is doing through a wall from ambient WiFi radiation raises legitimate concerns, and the research community has begun to address them directly, both technically and in terms of policy framing. Generalisation is another: most demonstrated systems work in the environments they were trained in, and performance drops when the geometry or hardware changes.

Neuralase sits within this broader field, working on the signal processing and machine learning problems that connect raw radio observations to meaningful human-centred outputs. The problems are hard. The potential applications, in health, safety, and human-computer interaction, are worth the effort.

This is not a mature technology. It is an active research frontier, and the distance between demonstrated capability and reliable deployed system remains significant. But the core physics is sound, the hardware is increasingly accessible, and the research output is accelerating. Radio-wave human sensing is becoming a real discipline.