World Sleep Industry ENCYCLOPEDIA

ballistocardiogram (BCG) non-contact monitoring modules

1.2.1

Name, Definition, and Module Boundary

A ballistocardiogram (BCG) non-contact monitoring module is an integrated sensing subassembly that detects minute whole-body reaction forces or motion associated with cardiac ejection and blood movement through a bed surface, mattress, bed frame, chair, or other support, then supplies signals related to mechanical heartbeats, respiration, and body movement to a host system. It may include one or more transducers, a mechanical carrier, analog conditioning, analog-to-digital conversion, a clock, embedded processing, and a communication interface. In this category, “non-contact” or “unobtrusive” usually means that no electrode must be adhered to the skin and no probe must be worn. It does not mean that the measurement has no mechanical contact through the support, requires no installation, or is necessarily imperceptible in every construction.

A module is not a complete monitoring device. It may expose only a raw waveform, filtered channels, candidate beat times, quality flags, or a limited rate estimate, while bedding compatibility, subject attribution, trend presentation, alarms, storage, and clinical interpretation remain responsibilities of the finished product. Entry 198, BCG mattress-type physiological monitoring devices, concerns a finished mattress-based structure, an overnight workflow, and user-facing results. This entry concerns the BCG signal source, transduction path, acquisition interface, algorithm boundary, and the integrator's responsibility for validation. An evaluation board or reference algorithm does not inherit the claims, safety, or reliability of a validated end product.

BCG names the measurement target and method rather than a single sensor material. Piezoelectric film, electromechanical film, strain gauges, load cells, accelerometers, pneumatic or hydraulic pads, microbend optical fibers, fiber Bragg gratings, and force sensors under bed legs can all support different implementations. A passive sensing element, an acquisition board with an analog front end, and a module with embedded estimation are not interchangeable deliverables. Procurement specifications should state the supplied processing level, supported mechanical placements, output definitions, timing behavior, calibration responsibility, and explicitly unsupported uses.

Physiological Mechanism and Waveform Meaning

When the heart contracts and accelerates blood into the great vessels, mass and momentum are redistributed within the body, producing small opposing forces, displacements, velocities, or accelerations at the body and its support. A recorded BCG is the observation of that mechanical sequence after transmission through tissue, posture, bedding, and the sensor structure. Shifts in center of mass caused by breathing, thoracoabdominal motion, and cardiorespiratory interaction are superimposed on it. A BCG is not the electrical potential measured by electrocardiography, the peripheral blood-volume change measured by photoplethysmography, or a direct measurement of blood flow, blood pressure, tidal volume, or oxygen saturation.

Classic longitudinal BCG nomenclature identifies waves such as H, I, J, and K, with the IJK complex associated with ventricular ejection and aortic blood flow. Modern bed sensors, however, use axes, constraints, transfer functions, and filters that differ from the early suspended or mobile beds on which this nomenclature was established. Peak direction, morphology, phase, and relative amplitude may therefore change. A local maximum in a contemporary waveform should not be labeled a classic J wave solely by appearance. If a module reports a J peak or another fiducial point, its definition, detection rule, latency, polarity convention, supported postures, and required signal quality should be documented.

Mechanical events in the BCG occur after myocardial electrical activation. The interval from an ECG R peak to a selected BCG feature contains electromechanical and ejection delays that can vary from beat to beat. BCG intervals can support heart-rate estimation, but they are not inherently identical to ECG RR intervals; small fiducial jitter, missed beats, and false detections can produce disproportionate errors in heart-rate-variability features. Inferring stroke volume, cardiac output, blood pressure, contractility, or disease from waveform amplitude or morphology requires separate validation for the particular device, population, posture, and reference method. The presence of an IJK-like pattern alone does not establish any of those quantities.

Transducers and Mechanical Coupling

Different transducers measure different physical quantities. Load cells and strain gauges can sense changes in support force, piezoelectric and electromechanical films preferentially respond to dynamic deformation, accelerometers measure local structural acceleration, and pneumatic, hydraulic, or optical arrangements first convert body action into pressure or an optical change. Their sensitivity, noise, static response, bandwidth, saturation, drift, and temperature dependence cannot be exchanged directly. A component data-sheet sensitivity is not the installed sensitivity, transfer function, or vital-sign performance of the assembled bed system.

The mechanical path from the person to the module determines whether a useful signal reaches the transducer. Mattress thickness and material, spring and foam damping, frame stiffness, bed-leg contact, mounting preload, sensor direction, body mass, height, posture, and position relative to the sensor all change the waveform. A module placed under a sheet, under a mattress, on a frame cross member, or beneath a bed leg is not being used in an equivalent configuration. Replacing the mattress, adding a thick topper, articulating an adjustable bed, moving a bed away from a wall, or changing floor contact can invalidate thresholds and models established for another mechanical path.

Turning, sitting up, scratching, speaking, coughing, snoring vibration, a pet, a bed partner, a fan, a washing machine, and building vibration can all exceed the cardiogenic component. When two people share a bed, their heartbeats and respiratory motions mix through the common structure; adding channels does not by itself guarantee correct source separation or subject attribution. Mechanical design should control sliding, resonances, structural impacts, cable motion, and inconsistent preload. The supported scope should state whether the module is intended for one person, two people, pets, adjustable beds, and specified bedding. When sources cannot be separated reliably, the system should report unknown attribution or invalid data rather than assign measurements to a person.

Acquisition Chain, Synchronization, and Interfaces

The acquisition chain must suit the transducer. A resistive bridge needs stable excitation and instrumentation amplification; a piezoelectric element needs an appropriate high-impedance or charge front end; an accelerometer requires stated range, noise density, axes, and mounting orientation. Every design must address bias, drift, common-mode pickup, power noise, electrostatic discharge, and overload protection. Respiration is a lower-frequency and often larger-amplitude component, whereas the mechanical heartbeat is faster and smaller. If analog filtering removes a needed band or clips its phase information before digitization, downstream software cannot reconstruct the original content.

Entering the bed or making a large movement can saturate the front end and ADC and can excite a decaying bed resonance. Interface specifications should disclose measurement range, sampling rate, effective resolution, bandwidth at each stage, filter phase or group delay, startup time, saturation recovery, noise, and fault states. A zero value may mean quiet conditions, cancellation after filtering, or a signal below the noise floor, but it can also mean an open or short circuit, loss of power, interrupted communication, or over-range handling. Zero output alone must not be interpreted as an empty bed, absent heartbeat, or absent breathing.

Clock synchronization is essential when multiple BCG channels, ECG, or another reference signal are combined. Each data unit should carry a timestamp, sequence number, unit, range, channel and axis identity, hardware and algorithm versions, and a validity indicator, with defined behavior for clock drift, buffering, packet loss, resampling, and reconnection. A module that supplies only an averaged rate should disclose its calculation window and update latency. Beat-to-beat analysis and multimodal event alignment require sufficiently raw or traceable data; the refresh time of a displayed number must not be treated as the physiological sample time.

Signal Processing, Outputs, and Failure Management

BCG is a nonstationary mixture of cardiac motion, respiration, voluntary or involuntary movement, and structural noise. Processing may include detrending, band separation, suppression of mains or mechanical interference, motion detection, candidate-peak extraction, template matching or autocorrelation, time-frequency analysis, channel selection, and signal-quality scoring. Fixed passbands do not guarantee perfect source separation. Respiratory harmonics may enter a cardiac band, heart and respiratory rates may approach integer relationships, and a mechanical resonance can create a stable spectral peak that is not physiological.

An algorithm should establish bed occupancy, supported subject count, completion of movement, adequate coupling, and sufficient waveform quality before producing a target output. During strong motion, clipping, poor signal-to-noise ratio, posture transitions, mixed-person signals, or markedly irregular rhythms, it should return invalid, low-confidence, or review-required status. It should not carry forward the last number, interpolate across an unsupported gap, or force the dominant spectral peak into a plausible vital sign. Relock time after movement, valid coverage, the longest gap, and reasons for failure are as important as mean error during clean segments.

Raw BCG, candidate mechanical beats, interbeat intervals, mean heart rate, respiratory rate, activity, bed exit, sleep stage, and respiratory events occupy different evidence levels. Accurate mean heart rate does not establish accurate HRV, arrhythmia identification, stroke volume, or blood pressure. Accurate respiratory rate does not establish airflow cessation, hypopnea type, oxygen desaturation, or an apnea-hypopnea index. BCG-based sleep staging also lacks direct electroencephalographic, electro-oculographic, and electromyographic observations. Each added output requires synchronous reference validation in the intended population and explicit handling of classes that cannot be determined.

Production, Calibration, and System Validation

Production testing should cover transducer offset and sensitivity, noise, bandwidth, polarity or axis orientation, interchannel crosstalk, analog gain, ADC behavior, clock accuracy, and communication. Mechanical calibration should use traceable force, displacement, or acceleration while controlling frequency, direction, point of application, preload, temperature, and fixture. A visible response to a hand tap only demonstrates that part of the chain responds; it does not show that heartbeats can be recovered under a supported mattress. When outputs already include algorithmic results, repeatable datasets and known expected results should also be used to confirm firmware, parameters, and model identity.

Reliability testing should address repeated loading, compression creep, mattress deformation, drop and impact, temperature and humidity cycling, liquid and cleaning-agent exposure, cable fatigue, connectors, adhesive aging, power variation, and long-duration drift. After exposure, verification should cover noise, gain, frequency response, saturation recovery, channel consistency, and fault detection in addition to basic startup. A change to the transducer, carrier, adhesive, front end, sampling rate, filters, or model can alter BCG morphology and should trigger risk-based revalidation rather than silent reuse of earlier coefficients.

Finished-system performance must be tested in production-representative construction and supported environments. Synchronized ECG is an appropriate reference for cardiac timing; respiratory results require a reference suitable for the claimed quantity, and sleep or diagnostic claims may require polysomnography or another clinical reference. Sampling should span expected body sizes, ages, postures, bedding, movements, abnormal rhythms, respiratory disorders, and single- and multi-person conditions. Reports should include error distributions, valid coverage, and failure cases rather than only correlations from motionless healthy volunteers. Passing a module bench test does not establish installed accuracy or clinical validity of the final device.

Safety, Compliance, Privacy, and Use Limits

The integrator remains responsible for electrical safety, electromagnetic compatibility, heat, liquid ingress, sharp edges, pinch hazards, bed stability, cable entanglement, flammability, cleaning, and material-contact risks. If software outputs participate in an alarm, sensor displacement, power failure, network loss, silencing, latency, and recovery paths also require verification. Regulatory identity follows the final intended use and claims. Classification of a ballistocardiograph in one jurisdiction does not automatically authorize an arbitrary consumer module, and describing a system as unobtrusive does not remove the need for end-product safety and performance evidence.

Bed micro-movements can reveal occupancy, bed exits, routines, activity, heart-rate trends, and respiratory trends, making the resulting records sensitive behavioral and health-related data. A connected product should minimize collection and protect device identity, configuration, transmission, storage, software updates, logs, and deletion rights. Local raw waveforms, derived indicators, and cloud-account data should be distinguished. Development debugging, model training, and service diagnostics also require a defined purpose, retention period, and access control; absence of a camera does not mean absence of privacy risk.

A BCG module cannot diagnose myocardial infarction, atrial fibrillation, sleep apnea, or sleep stage from a mechanical waveform alone, and it must not be assigned infant monitoring, life-support, nurse-call, or heartbeat-or-breathing-cessation alarm duties without dedicated validation and authorization. Chest pain, severe breathing difficulty, cyanosis, fainting, persistent palpitations, or unresponsiveness require action through the applicable emergency pathway; a user should not wait for module confirmation or use a normal trend to exclude disease. The finished product should document intended use, population, exclusions, invalid states, escalation behavior, and the consequences of unavailable data.