World Sleep Industry ENCYCLOPEDIA

Millimeter-Wave Vital Signs Radar

1.2.1

Name, Definition, and System Boundary

A millimeter-wave vital-signs radar is a complete device that illuminates a person with millimeter-wave radiofrequency energy and extracts respiratory and cardiac-related micromotion or human activity from changes in range, phase, amplitude, angle, or Doppler in the echo. It then reports vital-sign estimates or state information. The system typically includes the RF transmitter and receiver, antennas, clocks, sampling and signal processing, enclosure, power, firmware, and communications. A chip, antenna board, or component that requires further engineering is a monitoring module, not a ready-to-use radar product.

In telecommunications nomenclature, millimeter waves generally correspond to roughly 30–300 GHz, although the market may use the term loosely for nearby short-range radar bands. Purchasing and validation should rely on the actual operating band, bandwidth, transmission method, antenna configuration, and local authorization rather than the product name alone. Devices at 24 GHz, 60 GHz, 77 GHz, or another band can differ in propagation, resolution, regulation, and supporting evidence, so similar interfaces do not make their performance interchangeable.

In this category, vital signs usually means heart-rate, respiratory-rate, or related trends inferred from mechanical movement at the body surface. It does not mean direct measurement of cardiac electrical activity, airflow, oxygen saturation, blood pressure, or core temperature. Medical-device status depends on intended use, alerts, evidence, and regulatory status. Consumer radar, research prototypes, and clinical monitors must be described separately; words such as contactless, continuous, or artificial intelligence do not replace stated performance and conditions of use.

RF Chain and Meaning of Measurement

A single-frequency continuous-wave radar mainly senses micromotion through changes in phase or Doppler but generally lacks clear range separation for targets within the same beam. Frequency-modulated continuous-wave radar transmits repeated sweeps, estimates range from the beat frequency, and then analyzes phase changes within a selected range cell. Multiple-input multiple-output arrays can add angular information. Range, velocity, and angular resolution and frame rate depend on different parameters, so improving one does not prove improvement in the others or in physiological estimation.

Thoracoabdominal motion from breathing is generally stronger than mechanically generated cardiac micromotion, and the two can create harmonics, modulation, and spectral overlap. Radar phase represents composite displacement along the line of sight, not isolated movement of one anatomical point; bedding, multiple body surfaces, and reflection paths can all contribute. Separated respiratory and cardiac components are algorithmic results. Radar heart rate is not ECG, radar respiratory rate is not nasal or oral airflow, and waveform resemblance does not establish reliable detection of every R wave or respiratory event.

Antenna gain, beam width, polarization, array aperture, sweep bandwidth, phase noise, analog-to-digital conversion, and clock stability jointly limit signal quality. A stated maximum detection distance means that some target can be detected under specified conditions, not that weak cardiac motion remains accurate at that range. Raw point clouds, range maps, phase sequences, rate estimates, classifications, and alerts occupy different information levels. Documentation and interfaces should state how each is generated and when it is invalid.

Deployment Geometry and Bedroom Environment

Distance, mounting height, tilt, azimuth, thoracoabdominal orientation, and antenna field of view determine effective coverage. Side, prone, or back-facing posture, thick bedding, and movement away from the beam change both line-of-sight chest displacement and echo strength. Installation should follow the specific model instructions and be checked with usual postures, complete bedding, and the actual bed. A short bench trial or a motionless seated subject facing the radar does not establish full-night performance.

Walls, floors, metal frames, mirrors, heaters, furniture, and people create multipath reflections, while fan blades, curtains, air-conditioning flow, pets, robots, and activity at another bed introduce periodic or irregular motion. Multipath may strengthen an echo, cancel phase, or form a false target, so it is not uniformly beneficial or harmful. Deployment acceptance should separately test doors and windows, adjustable-bed motion, routine appliances, and caregiving activity and should record the fixed reflective background.

Relocating the device, loosening a mount, changing the room or bed position, and updating firmware parameters can invalidate the range-angle map. If the system calibrates an empty-room background, a bed zone, or a personal baseline, it should state when calibration must be repeated, how a wrong baseline is recognized, and how historical comparability is preserved. Fusion with a reference or another sensor also requires verified timestamps, clock drift, and event alignment. Similar time alone does not prove that two channels describe the same person or physiological event.

Motion, Multiple People, and Failure States

Turning, sitting up, coughing, talking, limb motion, and bed-frame vibration are usually much larger than cardiac micromotion and can cause phase jumps, migration between range cells, and a delay before relocking. The system should define motion detection, masking, reacquisition, and invalid-window rules rather than continuing to display smoothed values through strong motion. Accuracy calculated only during stationary windows omits the hardest periods, so performance reports must include valid coverage and recovery after movement.

Monitoring several people requires the system first to separate targets by range, angle, or other features and then extract micromotion for each. Echoes can occupy the same cell and mix when two people share similar range and angle, occlude each other, cross paths, share bedding, or move at the same time. Similar breathing rates and a breathing harmonic near another person's heart rate can also be misassigned. A multi-person claim should report separable geometry, supported count, identity continuity, side switching, and failure conditions rather than only two stationary targets in an open room.

The status interface should distinguish an empty room, a present target with invalid micromotion, unseparated multiple people, an obstructed radar, RF or processing failure, a network outage, and cloud delay. It should show the last successful estimate, signal quality, device orientation, firmware version, and local-buffer status. No reading does not mean zero heart or respiratory rate, and no alert does not prove no abnormality. When critical functions stop offline, that condition must be visible locally or at the receiving endpoint.

Algorithm Validation and Performance Reporting

Each output needs a matching synchronized reference: ECG for heart rate and interbeat intervals; a suitable respiratory belt, airflow channel, or clinical reference for respiratory rate; video or a calibrated sensor for body motion; and standard polysomnography for sleep stages and respiratory events. Validation must use the final RF hardware, antennas, enclosure, installation, firmware, and algorithm version. A motion simulator, stationary awake participants, or training data can test a limited link but cannot establish full-night home performance or performance in a patient population.

Continuous rates require bias, error distributions, agreement, window duration, and update latency rather than correlation or mean accuracy alone. Beat-level output, heart-rate variability, apnea, arrhythmia, and emergency alerts each require separate validation of temporal matching, sensitivity, specificity, false alerts, misses, and invalid periods. A stable average heart rate obtained from a long window does not prove that an instantaneous change, individual beat, or short pause is captured reliably.

Samples should cover intended ages, body sizes, respiratory and heart-rate ranges, postures, bedding, distance, angle, partners, and health states and disclose excluded windows, failed nights, training-test separation, funding, and conflicts of interest. A small single-center study or breath-hold trial in healthy adults cannot be generalized automatically to infants, older adults, arrhythmia, hypoventilation, assisted ventilation, or severe sleep-related breathing disorders. Performance should be reconfirmed after software changes, with model and parameter versions retained in longitudinal records.

RF, Electrical, and Coexistence Safety

RF safety and compliance must be evaluated for the actual band, output, antenna gain, waveform, duty cycle, distance, mounting orientation, and environment under the equipment-authorization and human-exposure requirements of the place of sale and deployment. Low power or contactless operation is not a substitute for compliance evidence. A report on one RF chip or module does not automatically cover a complete unit with a different antenna, enclosure, or amplifier. Users should not modify antennas, increase power, or move the product into an unevaluated near-body position.

Several radars, wireless networks, motion sensors, and other RF equipment in the same room may produce co-channel, adjacent-channel, intermodulation, or timing interference, while the radar itself must meet electromagnetic-compatibility and spectrum rules. Interference testing should include identical units side by side, different products in coexistence, restart behavior, and automatic channel or parameter changes rather than a quiet laboratory alone. When an anomaly occurs, configuration and logs should be preserved; required safety or medical equipment should not be disabled to improve the radar reading.

Mounts and power supplies should be protected against falling onto the bed, liquid ingress, blocked ventilation, trip hazards, and crushing by adjustable-bed mechanisms. Cleaning liquid must not enter the radome, ports, or vents, and a damaged enclosure, mount, or supply requires the manufacturer's process. Unusual heat, scorched odor, repeated restarts, cracking, looseness, or sudden output changes warrant stopping use. Repair, transfer, and disposal also require attention to calibration, RF configuration, accounts, and electronic waste.

Privacy, Alerts, and Medical Limits

Radar does not create an ordinary visible-light image, yet it can reveal room occupancy, location, movement, routines, partners, and cardiopulmonary trends. Range-angle tracks, point clouds, micromotion waveforms, and inferred labels are sensitive data. No camera does not mean no privacy risk. Documentation should define local versus cloud processing, retention, sharing, model-training use, and notice for other people in the room and should prevent data from guests or neighboring beds from being assigned to the wrong account.

A connected device needs proportionate authentication, role-based permissions, protected transmission, controlled updates, vulnerability communication, data export and deletion, and a stated support period. A factory reset should sever associations among the radar, gateway, application, cloud account, and third-party services. If timely alerts are offered, the trigger, recipient, latency, acknowledgment, and escalation chain must be explained and tested under internet and power loss, phone silencing, revoked permissions, server failure, and subscription expiry.

A consumer millimeter-wave vital-signs radar is not a substitute for life support, a nurse-call system, infant monitoring, or standard sleep diagnosis and cannot guarantee that an event will never be missed. Chest pain, marked breathing difficulty, cyanosis, fainting, seizure, or inability to awaken requires immediate action through local emergency procedures without waiting for radar confirmation. Persistent insomnia, daytime sleepiness, witnessed apneas, palpitations, or recurring abnormal trends should be assessed by a qualified professional using history, examination, and appropriate standard testing.