Falsifiable hypothesis
Research-stage radar sensing
Confidence-Qualified Physiological Sensing Through Complex Media
Complex Media Sensing investigates whether radar and inertial measurements can distinguish physiological micro-motion from interference caused by coat, fit, posture, and mounting movement.
The current research direction is canine wearable sensing. We propose to develop a method that reports an estimate when measurements support it and withholds one when they do not.
Human breadboard work has been conducted. Canine performance in the intended collar geometry remains to be established.
Development status · September 2026
What exists and what remains unproven
Based on founder reports, company materials, and source-code inspection; experiment outcomes were not independently verified for this summary.
| Area | Evidence status |
|---|---|
| Human breadboard work | Conducted as preliminary development; it does not establish canine performance. |
| Firmware and software | Source includes multi-bin fusion, motion gating, adaptive cancellation, and heuristic quality scoring. Source inspection does not establish hardware or canine validation. |
| Current engineering focus | Sensor-suite boards and communications testing. |
| Proposed estimator | Joint radar-inertial validity estimation and confidence calibration remain proposed research. |
| Canine performance | Not established in the intended collar geometry. |
| Beat-to-beat timing | Exploratory and separate from average cardiac-rate estimation. |
Research question
A periodic return is not automatically a physiological measurement
A collar radar measures motion relative to its sensor. Changing coat, fit, posture, and mounting can change that relationship. A periodic component may reflect physiology, artifact, or both; sensor and nearby tissue may also move together and obscure the relative motion of interest.
The key uncertainty is whether these features remain informative when sensor-to-body coupling changes. An IMU or more elaborate model cannot recover physiological information absent from the measurements.
Proposed method
Estimate, qualify, or abstain
Controlled displacement and media experiments will characterize how mounting conditions affect radar returns. Reference-labeled recordings will support development of a joint radar-inertial validity estimator. Its proposed confidence means the estimated probability that error falls within a predefined tolerance; calibration will be tested on held-out data.
Deployment inputs
Radar measurements and inertial context.
Research reference
Independent physiological reference labels support development and evaluation only.
Candidate output
A supported estimate with calibrated validity, or an explicit unavailable result.
Evidence-grounded comparison
| Method | Evidence status | Evaluation question |
|---|---|---|
| Radar-only estimate | Direct baseline for the proposed study | What does radar provide without inertial context? |
| Fixed motion gating | Published Acconeer human-breathing example pauses analysis during larger motion | Does a fixed threshold reject artifacts while retaining useful recording time? |
| Existing firmware heuristics | Firmware source includes multi-bin fusion, motion gating, adaptive cancellation, and heuristic quality scoring | Does the proposed method improve on current heuristic baselines? |
| Proposed joint radar-inertial validity estimator | Research hypothesis; not yet established | Can calibrated confidence improve error, false acceptance, and usable yield on held-out data? |
Scroll the comparison table horizontally to read every column.
Radar sensing and motion gating have prior work. Bowden et al. report canine radar physiology research; Acconeer documents motion-aware human breathing analysis. These sources provide context, not validation of the CMS collar or an exhaustive novelty review.
Proposed validation
Measure error, false acceptance, and usable yield together
Begin with controlled displacement phantoms and media analogs. Proceed to synchronized canine reference recordings only after required veterinary, animal-use, and assurance arrangements. No approved study institution or completed canine result is claimed.
Primary
Respiratory-rate estimation.
Secondary
Cardiac-rate feasibility.
Exploratory
Beat intervals, only if cardiac-source attribution is supported.
Compare held-out conditions, animals, and sessions. Report accepted-estimate error, false acceptance of corrupted windows, and usable-recording yield together; retain rejected periods and report reference failures separately. Freeze endpoint tolerances, thresholds, and test partitions before held-out evaluation. Poor calibration, low useful yield, or no improvement over simpler baselines would challenge the approach.
Controlled tests will include sensor-tissue co-motion. Canine recordings will begin at rest with repeated placement, documenting coat, posture, fit, and permitted movement. Each recording will retain hardware, firmware, configuration, timing, and analysis versions. All methods will use equivalent recordings, windows, and reference-quality rules.
Accepted-error risk is the fraction of accepted, reference-evaluable windows exceeding the endpoint tolerance. Corrupted-window acceptance instead uses all independently labeled corrupted windows as its denominator. Usable yield will state its eligible recording-time denominator. Numeric criteria remain protocol decisions, not measured results.
Application
An initial animal-health path, subject to evidence
A potential future application is animal-health monitoring with owner-authorized veterinary access to longitudinal records. This is a prospective use case; no product, partnership, or clinical benefit is claimed.
Value depends on sensing reliability, wearability, customer demand, and veterinary workflow usefulness, all still to be established. Working-dog and human uses are separate future directions requiring their own validation.
Earlier CMS-TM1 materials described a BARK-oriented human-canine concept with bench-only work in its proposed Phase I. That concept is historical; this update does not amend a prior submission or imply agency funding or endorsement.
Team and intellectual property
Research initiative of Lutz Innovations LLC
Complex Media Sensing is operated by Lutz Innovations LLC. Development experience includes embedded radar, firmware, mobile, cloud, and veterinary-workflow software. The proposed study still requires additional radar/DSP, statistical, and veterinary expertise.
Nathaniel Lutz reports filed patent applications relating to radar sensing. Patent applications do not establish validated performance or company ownership rights.
Research-stage technology. Canine collar performance and clinical utility remain to be established. No diagnosis, universal motion tolerance, radar oxygen-saturation measurement, or validated human medical use is claimed.
Contact
Research capability statement and inquiries
The contact button opens an email draft in your mail application; it does not submit or deliver a message by itself.
Complex Media Sensing is a research initiative of Lutz Innovations LLC.
Research-stage systems; not for diagnostic use.