# How Waves Can "See" People — Research for KAI Sensing / Biolock Security *What the research actually says about using waves (RF, radar, acoustic) to sense humans — presence, vitals, identity, even through walls — and how it maps to KAI as a security/biolock brain. Sourced; honest about limits. Compiled 2026-08-07.* --- ## 1. The physics — why a wave can sense a body at all A transmitted wave (radio, millimeter-wave, or sound) reflects off the human body. The body is never still: the **chest rises/falls with breathing (~millimeters)**, the **heartbeat flexes the chest surface (sub-millimeter)**, and limbs move when you walk. Each of these motions changes the reflected wave's **phase, frequency (Doppler shift), and time-of-flight**. Measure how the echo changes over time and you can reconstruct the motion that caused it — that's the whole trick. Two key terms you'll see: - **CSI (Channel State Information)** — in WiFi, the fine-grained amplitude+phase of each sub-carrier. A body moving through the room perturbs it measurably. - **Micro-Doppler** — the tiny velocity signatures of sub-motions (an arm swing, a leg's cadence, the chest wall). These fine signatures are what make *identification* possible, not just detection. --- ## 2. The modalities (the different "waves") - **WiFi sensing (CSI)** — uses ordinary 2.4/5 GHz WiFi. Cheap (even an ESP32), no new spectrum. Good for presence + coarse vitals; lower resolution. (He's said "not just WiFi" — so it's the floor, not the ceiling.) - **mmWave / FMCW radar (24–77 GHz, e.g. TI IWR6843, Google Soli 60 GHz)** — frequency-swept radar giving **range + Doppler**. The sweet spot: high-precision vitals, gait micro-Doppler, gestures, works in the dark, no camera (privacy-preserving), penetrates thin walls. - **UWB radar (ultra-wideband)** — very fine range resolution; strong for gait ID and precise ranging/presence. - **Acoustic / ultrasonic** — sound waves (often >20 kHz). Turns a mic/speaker into a sensor for gestures and even through-the-wall surveillance; short range, but nearly free hardware. --- ## 3. What waves can actually detect (the capabilities — "all the things") 1. **Presence, occupancy, count, and location.** Is someone in the room, how many, roughly where — from RF perturbations, no camera. 2. **Motion & activity recognition.** Walking, sitting, falling (big in elder-care), gestures. 3. **Vital signs, contactless, from across a room.** Breathing rate AND heart rate — even heart-rate *variability* — from sub-millimeter chest motion. mmWave FMCW radar does this to high precision; WiFi CSI does a coarser version. Multiple peer-reviewed results confirm it. 4. **Biometric identity from gait.** A person's walk has a unique **micro-Doppler signature**; deep-learning models identify *who* is walking from radar alone, including **open-set** ID (flag an unknown person, not just match a known list). This is the piece that turns "someone's here" into "*this specific person* is here." 5. **Through-wall sensing — and imaging.** RF passes through drywall. MIT's **RF-Pose** reconstructs a full human **skeleton/pose through a wall** from RF using a neural net (trained with a camera as teacher, then runs on RF alone). Ultrasonic through-wall surveillance exists too. Waves genuinely "see" people you can't see optically. 6. **Fine gesture recognition.** Google Soli (60 GHz radar) and ultrasonic active sensing read small hand gestures — sub-centimeter finger motion. --- ## 4. The security angle (your biolock / server use case) **Where waves HELP authentication:** - **Anti-spoof biometrics.** RF can beat photo/mask spoofing that fools cameras. Example: **RFace** uses COTS RFID to sense a face's **3D contour + material**, so a printed photo or flat mask fails — the wave "feels" depth and tissue, not just an image. - **Continuous / presence-based auth.** Instead of a one-time unlock, sense continuously and **auto-lock the instant the owner's signature leaves the room** — a strong *second* layer. - **Privacy-preserving.** No camera footage to leak; you store a signature, not a picture. **Where waves are a RISK (be honest for a *security* system):** - **Dual-use surveillance.** The same tech that lets KAI sense you lets an attacker sense you through a wall. RF sensing has a real "malicious exploitation" literature. - **Spoofing.** Radar motion sensors can be **spoofed with portable RF devices** injecting fake motion; gait/vitals are *harder* to fake than a photo but not unspoofable. - **Reliability drift.** Furniture moves, a second person enters, multipath changes — false accept/reject rates rise. A lock must **fail closed** (deny on uncertainty), log every decision, and keep a hardware override. **Verdict for a server biolock:** waves are excellent as a *factor* — presence, continuous-auth, anti-spoof depth sensing — but should **not be the sole gate.** Multi-factor + fail-closed. --- ## 5. How this maps to KAI (the fit is real) KAI is a **pattern-associative learning lattice** — which is exactly the brain this needs: - **Enroll:** feed the owner's sensor features (gait micro-Doppler, breathing/heart cadence, body reflection profile) → KAI encodes each into a 16,384-dim hypervector and stores it as a signature. - **Sense live:** stream features → KAI **resonance-matches** against the enrolled signature. - **Decide:** above threshold → owner present; open-set flag on an unknown → deny + alert. - **Adapt:** KAI's continual learning tracks slow drift (shoes, weight, room changes) without a full retrain. - KAI already leans this way in your own docs (PERSON-RECOGNITION-GOAL, the IR-camera sensory research) — this extends "sense the owner" from vision to RF. **Practical recommendation to start:** one **mmWave FMCW radar module (TI IWR6843 class, ~$150–300)** is the best first sensor — presence + vitals + gait micro-Doppler + gesture, no camera, works in the dark and through thin walls. Pipeline: radar → range-Doppler / micro-Doppler features (Python DSP) → stream into KAI → enroll/match → gate the lock as *one factor* alongside a credential, failing closed. --- ## Sources WiFi CSI sensing / vitals: [Non-Contact Heart Rate via WiFi CSI (PMC)](https://pmc.ncbi.nlm.nih.gov/articles/PMC11013971/) · [PhaseBeat CSI vital signs (Auburn)](https://www.eng.auburn.edu/~szm0001/papers/PhaseBeat_ACMHealth20.pdf) · [AI-enhanced CSI vital-sign review (PeerJ)](https://peerj.com/articles/cs-3375/) · [RuView ESP32 CSI presence+vitals](https://github.com/ruvnet/RuView) · [Synaptics Wi-Fi Sensing](https://www.synaptics.com/technology/wireless-technology/wi-fi-sensing) mmWave FMCW vitals: [Detection of vital signs based on mmWave radar (Nature Sci Reports)](https://www.nature.com/articles/s41598-025-09112-w) · [High-precision FMCW mmWave vitals (PMC)](https://pmc.ncbi.nlm.nih.gov/articles/PMC9572116/) · [Remote vital signs mmWave FMCW (Weizmann/Eldar)](https://www.weizmann.ac.il/math/yonina/sites/math.yonina/files/Remote_Monitoring_of_Human_Vital_Signs_Using_mm-Wave_FMCW_Radar.pdf) Gait micro-Doppler ID: [Human gait identification via UWB radar micro-Doppler (PubMed)](https://pubmed.ncbi.nlm.nih.gov/40039884/) · [Open-set human ID from gait radar](https://www.researchgate.net/publication/349459189_Open-Set_Human_Identification_Based_on_Gait_Radar_Micro-Doppler_Signatures) · [Deep transfer learning gait ID (IET)](https://digital-library.theiet.org/doi/10.1049/iet-rsn.2020.0183) Through-wall pose: [MIT RF-Pose project](https://rfpose.csail.mit.edu/) · [Through-Wall Human Pose Estimation (CVPR 2018 PDF)](https://openaccess.thecvf.com/content_cvpr_2018/papers/Zhao_Through-Wall_Human_Pose_CVPR_2018_paper.pdf) · [MIT AI sees through walls (CSO)](https://www.csoonline.com/article/565638/mits-ai-can-now-see-and-track-people-through-walls-using-wireless-signals.html) Acoustic/ultrasonic: [Ultrasonic through-the-wall surveillance](https://www.researchgate.net/publication/259614808_Ultrasonic_through-the-wall_surveillance_system) · [Ultrasonic active-sensing gesture recognition (arXiv)](https://arxiv.org/html/1712.00216v3) Security / spoofing: [RF Sensing Security & Malicious Exploitation survey (arXiv)](https://arxiv.org/html/2504.10969v1) · [RFace anti-spoof facial auth via RFID](https://www4.comp.polyu.edu.hk/~csyqzheng/papers/RFace-INFOCOM21.pdf) · [Spoofing attacks to radar motion sensors (NSF)](https://par.nsf.gov/servlets/purl/10230653)