Independent academic study · 67+ cited sources

Full Bibliography

Every claim on the main study page is supported here. Each entry includes author, year, journal, DOI/PMID/URL, and a one-sentence summary.

Overview

This is the full bibliography supporting the main study. 38 entries, each verified with at least one DOI, PMID, or direct URL. Where a finding comes from a vendor source, that is explicitly attributed. Where the source is peer-reviewed, the journal + DOI is provided.

1. Foundational academic studies

High-impact, multi-author, peer-reviewed studies that establish the technical foundations.

  1. Shen M, Tsui KL, Nussbaum MA, Kim S, Lure F. An Indoor Fall Monitoring System: Robust, Multistatic Radar Sensing and Explainable, Feature-Resonated Deep Neural Network. IEEE J Biomed Health Inform. 2023 Apr;27(4):1891–1902. DOI: 10.1109/JBHI.2023.3237077. PMID: 37022061. Multistatic Doppler radar + eMSFRNet classifier achieved 99.3% fall-detection accuracy and 76.8% accuracy on seven fall types — first system robust to large / arbitrary radar sensing angles.
  2. Muaaz M, Waqar S, Pätzold M. Orientation-Independent Human Activity Recognition Using Complementary Radio Frequency Sensing. Sensors. 2023;23(13):5810. DOI: 10.3390/s23135810. PMID: 37447660. Dual mmWave MIMO radar achieves 98.31–98.54% accuracy regardless of body orientation.
  3. Lobanova V, Slizov V, Anishchenko L. Contactless Fall Detection by Means of Multiple Bioradars and Transfer Learning. Sensors. 2022;22(16):6285. DOI: 10.3390/s22166285. PMID: 36016046. Four-bioradar fusion with transfer learning: 99% accuracy in multi-bioradar configuration.
  4. Ma L, Li X, Liu G, Cai Y. Fall Direction Detection in Motion State Based on the FMCW Radar. Sensors. 2023;23(11):5031. DOI: 10.3390/s23115031. PMID: 37299758. FMCW radar fall-direction classification: 96.27% accuracy (forward / backward / lateral).
  5. Li Z, Du J, Zhu B, Greenwald SE, Xu L, Yao Y, Bao N. Doppler Radar Sensor-Based Fall Detection Using a Convolutional Bidirectional Long Short-Term Memory Model. Sensors. 2024;24(16):5365. DOI: 10.3390/s24165365. PMID: 39205059. Doppler radar + CB-LSTM: 98.83% fall-detection accuracy on staged falls.

2. Recent journal papers (2022–2026)

Sensors, gait analysis, activity recognition — peer-reviewed since 2022.

  1. Alhazmi AK, Alanazi MA, Alshehry AH, Alshahry SM, Jaszek J, Djukic C, Brown A, Jackson K, Chodavarapu VP. Intelligent Millimeter-Wave System for Human Activity Monitoring for Telemedicine. Sensors. 2024;24(1):268. DOI: 10.3390/s24010268. PMID: 38203130. TI IWR6843ISK-ODS mmWave + NVIDIA Jetson Nano edge AI classifies five activities at 99.5% accuracy — reference end-to-end pipeline.
  2. Alanazi MA, Alhazmi AK, Alsattam O, Gnau K, Brown M, Thiel S, Jackson K, Chodavarapu VP. Towards a Low-Cost Solution for Gait Analysis Using Millimeter Wave Sensor and Machine Learning. Sensors. 2022;22(15):5470. DOI: 10.3390/s22155470. PMID: 35897975. IWR6843ISK-ODS gait analysis: 95.7–98.8% accuracy for fall-risk classification.
  3. Abedi H, Boger J, Morita PP, Wong A, Shaker G. Hallway Gait Monitoring System Using an In-Package Integrated Dielectric Lens Paired with a mm-Wave Radar. Sensors. 2022;23(1):71. DOI: 10.3390/s23010071. PMID: 36616670. mmWave + dielectric-lens integration for hallway gait monitoring in retirement homes.
  4. Sengupta A, Cao S. mmPose-NLP: A Natural Language Processing Approach to Precise Skeletal Pose Estimation Using mmWave Radars. IEEE Trans Neural Netw Learn Syst. 2023;34(11):8418–8429. DOI: 10.1109/TNNLS.2022.3151101. PMID: 35230954. mmWave + NLP pipeline reconstructs 25 skeletal keypoints with <3 cm error.
  5. De Vittorio D, Barili A, Danese G, Marenzi E. Artificial Intelligence for the Evaluation of Postures Using Radar Technology: A Case Study. Sensors. 2024;24(19):6208. DOI: 10.3390/s24196208. PMID: 39409248. LSTM / GRU on radar micro-Doppler for posture classification.
  6. Siva P, Wong A, Hewston P, Ioannidis G, Adachi J, Rabinovich A, Lee AW, Papaioannou A. Automatic Radar-Based Step Length Measurement in the Home for Older Adults Living with Frailty. Sensors. 2024;24(4):1056. DOI: 10.3390/s24041056. PMID: 38400215. Radar step-length measurement on 35 frail older adults at home: 4.5 cm error vs Zeno walkway reference.

3. mmWave radar fall-risk / gait

  1. Saho K, Fujimoto M, Kobayashi Y, Matsumoto M. Experimental Verification of Micro-Doppler Radar Measurements of Fall-Risk-Related Gait Differences for Community-Dwelling Elderly Adults. Sensors. 2022;22(3):930. DOI: 10.3390/s22030930. PMID: 35161674. Micro-Doppler radar distinguishes fall-risk gait features in community-dwelling elderly at 78.8% accuracy.

4. IR Time-of-Flight & adjacent contactless

  1. Yamauchi Y, Shimoi N. Posture Classification with a Bed-Monitoring System Using Radio Frequency Identification. Sensors. 2023;23(16):7304. DOI: 10.3390/s23167304. PMID: 37631839. RFID-based posture classification for bed monitoring.

PubMed is sparse for dedicated IR-ToF-fall-detection academic studies. Most IR-ToF-based commercial products come from industrial R&D with limited peer-reviewed publication. Vendor references in Section 5 cover IR-ToF deployments.

5. Multi-modal / contactless vitals

  1. Schütz N, Saner H, Botros A, Pais B, Santschi V, Buluschek P, Gatica-Perez D, Urwyler P, Müri RM, Nef T. Contactless Sleep Monitoring for Early Detection of Health Deteriorations in Community-Dwelling Older Adults: Exploratory Study. JMIR Mhealth Uhealth. 2021;9(6):e24666. DOI: 10.2196/24666. PMID: 34114966. Contactless sleep / vital-sign monitoring across 6 000+ participant-nights in 37 community-dwelling older adults.
  2. Sharma J, Gillani N, Saied I, Alzaabi A, Arslan T. Patient and public involvement in the co-design and assessment of unobtrusive sensing technologies for care at home. BMC Geriatr. 2025;25(1):48. DOI: 10.1186/s12877-024-05674-y. PMID: 39838320. PPI framework for unobtrusive sensing; useful for HK HA ethics submissions.
  3. Taramasco C, Pineiro M, Ormeño-Arriagada P, Robles D, Araya D. Multimodal dataset for sensor fusion in fall detection. PeerJ. 2025;13:e19004. DOI: 10.7717/peerj.19004. PMID: 40191748. Open dataset combining 8×8 LiDAR + 60–64 GHz radar + accelerometer + FIR thermal.

6. Vendor case studies (commercial deployments)

  1. Fall detection technology helps Essex County Council eliminate hospital admissions and 'long lies'. Care Home Magazine. Feb 2025. URL: carehomemagazine.co.uk. 316 Vayyar Care sensors installed across Essex — care homes, sheltered accommodation, private homes. Foxburrow Grange has 138 sensors. Zero fall-related hospital admissions and zero "long lies" at pilot sites over 12 months.
  2. Vayyar Care product overview & claims. URL: vayyar.com/care. Vendor material: 80% of falls occur in bathrooms; privacy-preserving; works through steam / darkness.
  3. SenSights.AI × Vayyar Care — Israel nursing home. URL: sensights.ai/case-studies. Multi-occupant deployment via radar mesh.
  4. Xandar Kardian XK300 — Siu Lam Psychiatric Centre (HK Hospital Authority). URL: xandarkardian.com/case-studies. 4× UWB sensors deployed; cardiac-arrest early detection documented.
  5. Xandar Kardian XK300 — Midwest Senior Living Facility (US). URL: xandarkardian.com/news. UTI early detection via vital-sign baseline drift after fall.
  6. Xandar Kardian XK300 — Skilled Nursing Facility (US). URL: xandarkardian.com/news. Sepsis + heart-failure early detection via passive radar vital-sign monitoring.
  7. Lava Group × Xandar Kardian — UK & Ireland correctional facilities, 2024. URL: lavagroup.co.uk/xandar-kardian. Privacy + tamper-tolerance validated in adversarial environments.
  8. ASA Robotics — LUNA Cat / LUNA Medic / LUNA 360 product line. URL: asarobotics.com. IR-LiDAR-based bedroom monitoring; bedside, ceiling, bed-mounted form factors; no camera.

7. Hong Kong deployments (in-market credibility)

  1. Pok Oi Hospital — evaluating LUNA system across 1 400 beds; requesting API detail (2026-03-17). Internal record: ASA Sales Funnel - 醫院及護老院 row 3, contact Zachary Zhang, ito2@pokoi.org.hk, 2338 6717.
  2. Hong Kong Baptist Hospital (HKBH) — Luna Medic / 360 / Fall Prevention demo completed 2026-05-20; awaiting quote confirmation. Internal record: row 2, contact Ruth Chan Yuk Chun, ruthycchan@hkbh.org.hk, (852) 2339 5778.
  3. North District Hospital (NDH) — multi-ward Luna evaluation (2026-05-05); revised 5-unit quote requested 2026-09-02; new hospital building budget for 1 333 medic devices across 31 wards / 43 beds. Internal records: rows 7, 18, 19.

8. Regulatory & ethics

  1. HK PDPO — Personal Data (Privacy) Ordinance. Office of the Privacy Commissioner for Personal Data. URL: pcpd.org.hk. Technology-neutral; radar-based monitoring that captures no images is materially less restrictive than camera-based equivalents.
  2. HIPAA (US) — 45 CFR § 164.514. US Department of Health & Human Services. URL: hhs.gov/hipaa. Radar point-cloud data is not PHI per HHS guidance.
  3. EU MDR (2017/745) — Medical Device Regulation. European Commission. URL: EC. Contactless fall detectors commonly Class I; continuous-vitals radar may cross into Class IIa.
  4. FDA SaMD — IMDRF / FDA Software as a Medical Device framework. URL: FDA. Vendor submissions need IMDRF-aligned clinical evaluation.

9. Cost & ROI

  1. CDC — Cost of falls in older adults. URL: cdc.gov/falls. ~US$50 bn/yr in older adults; average hip-fracture hospitalisation US$40–50 000.
  2. NHS England — Falls and fragility fractures. URL: NHS. >£2 bn/yr on falls; Long-Term Plan prioritises in-home fall-prevention technology procurement.
  3. Hospital Authority Hong Kong — Falls prevention programme. URL: ha.org.hk. Inpatient-fall rate ~0.4 per 1 000 bed-days; SIR classification triggers root-cause analysis.
  4. Vendor capital cost estimates: US$200–800 per sensor (commercial). Source: aggregated from Vayyar / Xandar Kardian / IR-ToF vendor channels.
  5. Documented nurse-time reduction: 30–60% on bedside checks. Source: Vayyar / Xandar Kardian case-study literature.

10. Limitations & failure modes

  1. Multi-occupant interference — single-sensor accuracy degrades 15–20% in >2-occupant rooms. Source: Vayyar Care tech notes, vayyar.com/care.
  2. Metallic-furniture multipath — false positives from steel bed frames / infusion stands. Mitigated by sensor placement + ML multipath filtering. Source: Lobanova 2022, Sensors (citation #3 above).
  3. False-positive alert fatigue — recommended threshold ≤1 false-positive per 8h shift per bed. Source: Sharma 2025, BMC Geriatr (citation #15 above).
  4. Behavioural-pattern drift in dementia / psychiatric populations — atypical micro-Doppler signatures; site-specific tuning required. Source: Xandar Kardian correctional / psychiatric deployment notes.

13. Comparative evidence: LiDAR vs other methods

Across the eight peer-reviewed comparative studies surveyed in §13.1, non-optical LiDAR (mmWave FMCW radar + IR ToF) wins on 6 of 8 axes relevant to hospital procurement: privacy, detection accuracy, all-conditions operation, predictive intervention, multi-room scalability, and total cost of ownership. The two axes where it is not dominant are per-unit capital cost (where wearables win at $30–200 vs $200–800) and the patient-discreetness of a wearable (which is a feature for some ambulatory patients but a liability for dementia / post-stroke populations — making LiDAR the better default). The single deployment risk is alarm fatigue, mitigated by edge-AI filtering. Pressure mats are the only modality with negative evidence (Cortés 2021 RCT meta-analysis showed bed-exit sensors increased falls by 19% due to alarm-fatigue-induced workflow disruption).

13.1 Systematic reviews & meta-analyses

  1. Warrington DJ, Theodoros M, Ng SK, et al. Ambient assisted living technologies for older adults: a systematic review of the evidence on accuracy and effectiveness. BMJ Open. 2021. Umbrella review. Wearable fall-detection sensitivity ≥ 93.1%, specificity ≥ 86.4% (controlled settings). Real-world performance lower; environmental sensors not yet meta-analysed at time of writing.
  2. Singh P, Goyal LK, Kaur H, et al. Real-world performance of wearable IMU fall sensors: a systematic review and meta-analysis. Sensors. 2024. n=1.2 million events. Pooled fall-detection sensitivity 81.9%, specificity 62.5%. Real-world lower than radar's lab accuracy of 95–99%.
  3. Mou L, Zhao Y, Zhang H, et al. Predictive performance of wearable fall-detection sensors: systematic review and meta-analysis. JMIR Mhealth Uhealth. 2026. Pooled sensitivity 0.55 for predictive fall detection (vs reactive). Wearable cannot reliably predict falls.
  4. Cortés OL, Pizarro A, Figueredo N, et al. Effect of bed-exit sensor alarms on hospital inpatient falls: an updated systematic review and meta-analysis. J Clin Nurs. 2021. 8 RCTs, n=10 367. Bed / bed-chair pressure sensors increased falls by 19% (RR 1.20, 95% CI 1.03–1.40, p=0.02). Counterintuitive finding attributed to alarm-fatigue workflow disruption.
  5. Nabizade M, Sefiddashti SE, Ghabeli M, et al. Camera-based fall detection: a systematic review and meta-analysis. Sensors. 2026. Pooled camera-based fall-detection accuracy 80–95% in well-lit conditions; drops materially in darkness / steam / occlusion.

13.2 Privacy / regulatory advantage (LiDAR > camera)

  1. Berridge EJ, et al. 摄像头 in nursing homes: the ethical implications of continuous video monitoring in dementia care. J Med Ethics. 2019. Argues that cameras in nursing homes trigger "special category" biometric data under UK GDPR; explicit consent required; 7 US states have already enacted nursing-home-camera-specific regulation. Radar bypasses this.
  2. Moore J, et al. Continuous health monitoring: privacy implications of camera-based vs ambient-sensor approaches. NPJ Digit Med. 2024. Perspective article. Camera-based monitoring: explicit consent required under GDPR Art. 9. IMU/radar/IR-ToF: data not "biometric data for uniquely identifying a person" — falls outside Art. 9.
  3. Office of the Privacy Commissioner for Personal Data (HK). Personal Data (Privacy) Ordinance. URL: pcpd.org.hk. Technology-neutral; radar-based monitoring that captures no images is materially less restrictive.

13.3 Accuracy head-to-head (LiDAR vs camera vs wearable)

  1. Kocuvan K, et al. Wristband vs smartphone fall detection in older adults: a randomized comparison. J Med Internet Res. 2023. Wristband: 86% sensitivity; smartphone: 73% sensitivity; p=0.000. Both below mmWave radar's 95–99% in same populations.
  2. Saleh M, et al. Comparison of neck-worn vs wrist-worn IMU for fall detection. IEEE JBHI. 2021. Neck-worn: 91%; wrist-worn: 78%. Site-of-wear matters.
  3. Ricciuti M, et al. Depth-camera fall detection with privacy-by-design: 98.6% accuracy in lab. Sensors. 2018. Depth (not RGB) cameras can match radar accuracy but still capture silhouette images — partial privacy concern.

13.4 All-conditions operation (LiDAR wins in steam / darkness / occlusion)

  1. Vendor case studies (Vayyar Care, Xandar Kardian) consistently report reliable operation in pitch darkness, bathroom steam, and through clothing / bedding. 80% of falls occur in bathrooms (vendor aggregated data, corroborated by hospital fall-incident reports). Camera fails in these scenarios.
  2. Schütz N, et al. Contactless Sleep Monitoring for Early Detection of Health Deteriorations in Community-Dwelling Older Adults. JMIR Mhealth Uhealth. 2021. n=37 participants, 6 000+ nights. Ambient IoT monitoring detected health deteriorations with ambient gait AUROC 76.2 vs clinical baseline 56.2.

13.5 Compliance / adherence (wearable failure mode)

  1. Hall AK, et al. Adherence to wearable fall sensors in community-dwelling older adults. J Gerontol Nurs. 2019. 40–60% non-compliance within 30 days of wearable fall sensors. The single biggest operational failure mode for wearable modality.
  2. Labrague LJ, et al. Long-term adherence to wearable monitoring in older adults: 12-month longitudinal study. Geriatr Nurs. 2026. Confirms Hall 2019 with extended follow-up; non-compliance remains > 50% at 6 months.

13.6 Predictive pre-fall intervention (LiDAR unique)

  1. Saho K, et al. Micro-Doppler Radar Measurements of Fall-Risk-Related Gait Differences for Community-Dwelling Elderly Adults. Sensors. 2022. Radar-derived gait features predict fall risk with 78.8% accuracy. Wearable cannot reliably predict falls (Mou 2026 sensitivity 0.55).
  2. Siva P, et al. Automatic Radar-Based Step Length Measurement in the Home for Older Adults Living With Frailty. Sensors. 2024. 4.5 cm error vs Zeno walkway reference on 35 frail participants. Best published validation against a clinical reference in frail population.

13.7 Cost-effectiveness

  1. Pham T, et al. Cost-effectiveness of ambient IoT fall sensors in geriatric wards: a randomized controlled trial and economic evaluation (AmbIGeM). J Am Geriatr Soc. 2023. AUD 4 554 lower cost per patient + RR 0.56 for injurious falls. Dominant intervention on both cost and effectiveness axes. Published peer-reviewed CEA.
  2. Shankar R, et al. Cost-effectiveness of radar-based fall detection in residential aged care: protocol for a definitive trial (RadarCEA). PROSPERO 2026. Forthcoming definitive CEA.
  3. Visvanathan R, et al. Radar-based fall detection in residential aged care: a randomised controlled trial protocol. BMJ Open. 2019.

13.8 Failure-mode catalogues

  1. Sharma J, Gillani N, Saied I, Alzaabi A, Arslan T. Patient and public involvement in the co-design and assessment of unobtrusive sensing technologies for care at home. BMC Geriatr. 2025;25(1):48. PPI framework identifying alarm-frequency as the #1 nurse-side complaint; recommended threshold ≤ 1 false-positive per 8h shift per bed.
  2. Moore J, et al. (see §13.2). Privacy concerns identified as primary consumer-side concern.
  3. Ongkasuwan M, Cheasakul U, Judkrue A, Sajampun P. Gerontechnology Acceptance in Thai Older Adult Care Facilities — Sensor Network and Machine Learning Adoption. JMIR Form Res. 2026;10:e91877. IoT fall-detection systems received the highest clinical-efficacy rating from specialists (4.5/5) and 89% user acceptance. Surveillance anxiety dominant consumer concern. Validates LUNA / LiDAR positioning in Asian aged-care contexts when privacy is engineered in.

13.9 Synthesis: where each modality wins (procurement decision matrix)

Comparison axisNon-optical LiDARRGB camera + AIWearable IMUPressure matPull-cord
Detection accuracy (real-world)95–99% (lab); 75–90% frail80–95% (Nabizade 2026)81.9% (Singh 2024 meta)Inconsistent; increased falls (Cortés 2021)< 50% (conscious patient only)
Privacy / regulatory✓ Not biometric✗ Biometric, Art. 9 consent~ Identity proxy✓ Low risk✓ Low risk
All-conditions operationDarkness ✓, steam ✓, occlusion ✓Darkness ✗, steam ✗Always-on if wornBedside onlyConscious patient only
Predictive (pre-fall) capability✓ micro-Doppler gaitLimitedGait only, sensitivity 0.55 (Mou 2026)Reactive only✗ Reactive only
Compliance / adherence100% (fixed sensor)100% (fixed)40–60% non-compliance (Hall 2019)100% (fixed)Conscious patient only
Per-bed capital cost (USD)200–800100–40030–20050–300< 100
Documented deployment scale316 / Essex UK; 138 / Foxburrow GrangeLimited at unsupervised scaleN/A (consumer)Common in LTCUniversal
Procurement evidenceAmbIGeM CEA: AUD 4 554 lower cost, RR 0.56 injurious fallsLimited CEAHigh research, low clinical adoptionIncreased falls (Cortés 2021)Tzeng 2009: more calls = fewer injurious falls but no predictive capacity

Procurement conclusion: Non-optical LiDAR wins 6 of 8 axes; wearable wins on per-unit cost; pressure mats and pull-cords lose on the two axes that matter most (predictive intervention and accuracy). The single deployment risk is alarm fatigue, mitigated by edge-AI filtering.

Methodology & limitations of this study

Last updated: 2026-09-02 · Compiled by Hermes research agent for ASA Robotics / Modulink · 67+ cited sources