AI Through-Wall Detection Explained

For decades, "seeing through walls" was a capability reserved for science fiction movies or ultra-expensive military hardware. In movies, through-wall radar is often depicted as showing a high-resolution, full-color 3D model of a person on a handheld screen.

In reality, through-wall sensing does not produce optical-quality images. Instead, it relies on interpreting electromagnetic signals that bounce off moving bodies and return through physical barriers. By applying modern artificial intelligence to these returned signals, systems like RuView can extract extremely accurate classifications of human presence, walking patterns, and breathing rates—even if the person is behind a thick drywall partition.

In this article, we will explore the physics of electromagnetic penetration, look at how different building materials affect Wi-Fi signals, and explain how RuView's Rust engine processes these through-wall perturbations.

To understand how Wi-Fi signals pass through walls, we must look at how electromagnetic waves interact with matter. When a radio wave strikes a barrier (like a wall), three things happen: a portion of the wave is **reflected**, a portion is **absorbed** (turned into heat), and a portion is **transmitted** (passes through).

The ratio of reflection, absorption, and transmission depends on the wave's frequency and the material's physical properties—specifically its electrical conductivity and permittivity.

Common building materials affect Wi-Fi signals (2.4 GHz and 5 GHz) differently:

• **Drywall & Wood**: These materials have low electrical conductivity and low water content. At 2.4 GHz, they are highly transparent, absorbing less than 1 to 2 dB of signal strength.
• **Brick & Concrete**: These materials are much denser and contain mineral water. They attenuate signals more heavily (typically 6 to 15 dB of loss), reducing the range of through-wall sensing but still allowing transmission.
• **Metal & Reinforced Concrete**: Steel-reinforced concrete reflects nearly 100% of Wi-Fi waves. Sensing through solid metal or heavy rebar meshes is extremely difficult.

Because standard interior walls in homes and offices are primarily made of wood and drywall, they are largely transparent to Wi-Fi waves. This allows RuView to easily track human presence between rooms. For details on picking the right hardware to run these sensors, see the Best ESP32 Boards guide.

When a Wi-Fi signal travels through a wall, reflects off a human, and travels back through the wall to the receiver, the raw signal contains reflections from both the static wall and the moving human. In fact, the reflection from the wall is often thousands of times stronger than the reflection from the human. This is known as the "near-wall flash" problem.

To isolate the tiny human signal, RuView applies advanced digital signal processing:

1. **Static Clutter Suppression**: Since the wall does not move, its reflection is constant. RuView's Rust engine tracks the time-average of the CSI matrix and subtracts it from each incoming packet. This removes all static reflections (walls, doors, tables), leaving only the dynamic components.

2. **Doppler Processing**: As the human body moves, it shifts the frequency of the reflected waves. By computing the Doppler frequency shift, RuView can identify the speed and direction of the human relative to the wall, ignoring the static barrier completely.

For a full breakdown of this signal pipeline, read How WiFi CSI Works.

Because walls introduce complex distortions (refraction and phase delays), traditional mathematical models often struggle to generalize across different rooms. This is where AI becomes essential.

RuView uses deep neural networks (specifically Convolutional Neural Networks and LSTMs) trained on diverse datasets captured in different architectural layouts. The network learns to recognize the invariant signatures of human movement, such as the specific frequency patterns associated with walking or falling, regardless of the wall material or room geometry.

Running these models locally at the edge ensures low latency and absolute privacy. There are no video files to hack, and no raw imagery is ever generated. Explore this architecture on the How It Works Page or view our complete features list on the Features Page.

While through-wall sensing is incredibly powerful, it has physical limits:

• **Range Limits**: Each wall a signal penetrates reduces the overall range. While open-space range can exceed 20 meters, through-wall range is typically limited to 5-10 meters.
• **Vitals Limitations**: Breathing detection through walls requires the subject to remain relatively still. Moving walls (e.g. thin partitions vibrating from HVAC wind) can introduce interference.

Despite these limits, RuView is a highly robust solution for privacy-friendly occupancy and fall detection. If you have questions about deploying it in your space, check out our FAQ Page or get in Contact with us.