LampTide · drawing set · v1.0
Part I · Cover Docket · Abstract · v1.0
- Hue / LIFX / Nanoleaf — manual or scene-based, foreground notification
- Calm Automaton (CHI EA '17) — DIY toolkit, not physiological-coupled
- Towards Calm Displays (CHI '17) — ambient-reflective, no occupant-state coupling
- MoodLight (CSCW '15) — biosensor → hue, but foreground & no perceptual floor
- DeLight (PUC '18) — HRV biofeedback coaching, foreground loop
- Apple Focus mode — UI mode, not ambient
- Circadian-lighting — clock-driven, not state-driven
- Sheet 1 FIG. 1 Scene · desk lamp breathing with occupant 100
- Sheet 2 FIG. 2 24-h calm-index → bulb-intensity trace 200
- Sheet 3 FIG. 3 Functional block diagram · sensor → driver 300
- Sheet 4 FIG. 4 Perceptual-threshold floor (period vs depth) 400
- Sheet 5 FIG. 5 Prior-art comparison matrix 500
- Sheet 6 FIG. 6 Hardware-embodiment taxonomy 600
- Sheet 7 FIG. 7 User-foreground override · state diagram 700
- Sheet 8 FIG. 8 Multi-room mesh extension · per-room inference 800
- Sheet 9 FIG. 9 Per-occupant θ calibration · adaptive convergence 900
- 100sphysical scene (FIG. 1)
- 200s24-h trace events (FIG. 2)
- 300spipeline blocks (FIG. 3)
- 400sperceptual-threshold map (FIG. 4)
- 500sprior-art comparison (FIG. 5)
- 600shardware embodiments (FIG. 6)
- 700soverride state machine (FIG. 7)
- 800smulti-room mesh (FIG. 8)
- 900scalibration loop (FIG. 9)
Part II · Drawings FIG. 1 – 9 · Sheets 1 – 9
Part III · Specification Field · Background · Summary · Brief description · Detailed description · Alternatives · Robustness
- H05B 47/115 · smart luminaires
- H05B 47/19 · control by sensor
- G06F 1/16 · portable computing
- F21V 23/04 · fixture electronics
- G16H 40/63 · health monitoring
Smart luminaires (Philips Hue, LIFX, Nanoleaf) are widely deployed but operate predominantly as foreground notification surfaces or as manually-toggled scene controllers. Circadian-lighting products (e.g. Apple Adaptive Brightness, dedicated F.lux derivatives) modulate color temperature on a clock-driven schedule independent of occupant state.
The academic literature on calm technology (Weiser & Brown 1995; Case 2015) and ambient information displays (Mankoff et al. 2003; Pousman & Stasko 2006) argues for peripheral displays that fade into background attention. Calm Automaton (CHI EA 2017) demonstrates a DIY toolkit; Towards Calm Displays (CHI 2017) demonstrates ambient-reflective matching of bedroom illumination. None couple the ambient output to a locally-inferred occupant physiological state.
The disclosed system supplies that coupling, drawing the occupant-state signal from the RoomMirror parent specification (passive Wi-Fi CSI) or from a standalone phone-IMU + microphone baseline. The novel hop is the peripheral-attention-bounded mapping from physiological state to luminaire setpoint, with the modulation period and depth held strictly below a perceptual-threshold floor (FIG. 4 · 402) so that the bulb never enters the foreground of attention.
An occupant-state inference module (302) operating locally within a dwelling produces a calm index. The calm index is smoothed by a low-pass filter (304) with cutoff period bounded below by a perceptual-threshold floor (≥ 30 s, per FIG. 4). The smoothed calm index is mapped by a luminaire driver (306) to brightness and correlated color temperature setpoints of a target bulb (308), such that the temporal modulation of the bulb is by construction peripheral to occupant attention rather than foreground.
The mapping is a configurable look-up table; the period of the resulting visible modulation lies in the range 30–300 s; the depth (Δ% brightness, ΔK temperature) lies below the perceptual-threshold curve of FIG. 4. No occupant identifier, location, or raw sensor stream is transmitted off-host.
- FIG. 1A scene view of luminaire 102, occupant 104, and respiration trace 106; glow lobes around the bulb represent slow intensity modulation tracking the calm index.
- FIG. 2A 24-hour trace of the calm index 208 and the phase-lagged bulb intensity 206; annotated bands 202 (deep work) and 204 (wind-down) demonstrate envelope behavior.
- FIG. 3Functional block diagram showing two alternative sensor stacks (CSI receiver 301a and phone IMU + mic 301b) feeding occupant-state inference 302, calm-index smoothing 304, luminaire driver 306, and bulb 308.
- FIG. 4A period-versus-depth phase diagram with the perceptual-threshold floor 402 shaded; the LampTide operating envelope 406 sits strictly below the curve, with operating point 408 marked.
- FIG. 5Prior-art comparison matrix 502–520; LampTide 520 is the unique row satisfying all five criteria.
- FIG. 6Hardware-embodiment taxonomy 602–608; primary single-bulb embodiment in accent, alternatives 604 (mesh), 606 (panel), 608 (smart blind) share the perceptual-floor constraint.
- FIG. 7Two-state machine NORMAL 702 ↔ HOLD 704 visualising the user-foreground override of claim 4; transitions 706 (override asserted) and 708 (release + N-second cooldown).
- FIG. 8Multi-room mesh extension 802; three independent per-room pipelines 808a/b/c each driving a luminaire 810a/b/c, with optional inter-room sync 816 confined to the local LAN.
- FIG. 9Per-occupant calibration loop INIT 902 → MODULATE 904 → SENSE 906 → ADJUST 908 → loop 910, with three example occupant trajectories 912 converging to personal operating depths over ~30 days.
The local computing device of FIG. 3 · 300 may comprise, in some embodiments, a small single-board computer (e.g. a Raspberry Pi or equivalent) co-located with the luminaire; in some embodiments, a smart-home hub running a local agent; in some embodiments, a portable computing device of the occupant operating in a low-power background mode.
The passive sensor stack may comprise either (a) a Wi-Fi channel-state-information receiver per the RoomMirror parent disclosure (block 301a, FIG. 3), which infers respiration and gross motion from the room-scale channel impulse response without imaging the occupant, or (b) a portable computing device's inertial-measurement-unit and microphone (301b), which infers an analogous occupant-activity envelope from on-body and acoustic correlates of presence. In some embodiments both sensor stacks are present and the inference module 302 fuses them by weighted average.
The calm-index smoothing 304 is a low-pass filter with cutoff period τ_p ≥ 30 s. The cutoff is configurable but bounded below by the perceptual-threshold floor of FIG. 4 · 402, which encodes the joint period × depth boundary above which the modulation would be perceived as a foreground flicker. The threshold curve is derived from classical flicker-fusion and brightness-discrimination psychophysics (Kelly 1961; De Lange 1958) and from contemporary measurements of just-noticeable difference in CCT (Wei et al. 2014).
The luminaire driver 306 maps the smoothed calm index to a setpoint pair ⟨brightness, CCT⟩. The mapping is a configurable look-up table; in some embodiments it is replaced by a learned function fitted to the occupant's own historical preferences. In some embodiments only brightness is modulated; in some embodiments only CCT; in some embodiments both.
The target bulb 308 may comprise a single smart luminaire (e.g. Philips Hue, LIFX, or Nanoleaf); in some embodiments, a mesh of co-located luminaires actuated in synchrony; in some embodiments, a non-bulb ambient surface (e.g. a Nanoleaf panel, a smart blind, or a low-resolution e-ink ambient frame).
A user-foreground override (claim 4) is asserted when the local computing device detects that the occupant has entered a foreground-attention state with respect to the luminaire itself — for example by an explicit gesture, by a voice command, or by a sustained gaze fixation on the luminaire as inferred from a webcam (see Hong et al. MobileHCI 2023 for sustained-gaze trigger precedent). Upon override the modulation enters a hold state at the most-recent setpoint, as depicted in FIG. 7.
The on-device-only constraint of claim 1 is supported by two converging facts of the present art: first, the privacy threat posed by even encrypted smart-home traffic (Apthorpe, Reisman & Feamster 2017) makes any off-host transmission of occupant signals risky regardless of encryption; second, contemporary tiny-inference benchmarks (Banbury et al. MLPerf Tiny MLSys 2021; Ren et al. TinyML survey 2023) demonstrate that the inference module 302 and the smoothing filter 304 fit comfortably within the compute and memory envelope of either a bulb-class microcontroller or the low-power background mode of a portable computing device. The wherein-clause is therefore enabling, not aspirational.
- Output surface. In some embodiments the target bulb (FIG. 3 · 308) is replaced by a Nanoleaf-style panel, a smart blind that modulates daylight transmittance, or a low-resolution e-ink ambient frame — in all cases preserving the slow-cycle / sub-perceptual constraint.
- Sensor fusion. In some embodiments the CSI receiver (301a) and phone IMU+mic (301b) operate concurrently and their respective occupant-state estimates are fused by weighted average or by a learned gating function.
- Modulation axis. In some embodiments only brightness is modulated; in some embodiments only CCT; in some embodiments both; in some embodiments additionally the spatial distribution of multiple co-located luminaires.
- Smoothing filter. In some embodiments the low-pass smoothing (304) is a single-pole IIR with τ_p = 60 s; in some embodiments a Kalman filter with a state model of occupant respiration; in some embodiments an adaptive filter whose cutoff tracks the dominant time-scale of the inferred occupant activity.
- Mapping function. In some embodiments the look-up table (claim 1 · (d)) is replaced by a learned function trained on the occupant's own historical ⟨state, setpoint⟩ preferences.
- Override semantics. In some embodiments the user-foreground override (claim 4) is gesture-triggered; in some embodiments voice-triggered; in some embodiments triggered by sustained gaze fixation on the luminaire.
- Local trace logging. In some embodiments the system additionally persists the calm-index trace and luminaire setpoints locally for the occupant's own later review, without transmitting them off-host.
- Multi-room mesh. In some embodiments multiple instances of the disclosed pipeline are deployed across rooms of the same dwelling (FIG. 8 · 802); per-room inference 808a–c is independent, and rooms may optionally synchronise setpoints via local-LAN traffic 816, with no traffic transiting the dwelling boundary.
- Sensor unavailable. If the CSI receiver (301a) is unavailable — e.g. the Wi-Fi router is replaced with one whose firmware does not expose channel-state-information — the system falls back to the phone IMU + mic stack (301b). If both stacks are absent, the luminaire driver (306) freezes at the most-recent setpoint (analogous to HOLD 704) and an unobtrusive indication is surfaced to the occupant.
- Multiple occupants. When two or more occupants are detected in a single room, the calm-index 208 is taken as the running average of their inferred states, weighted by inferred proximity to the luminaire. In some embodiments the system enters a per-occupant mode if individual occupants are reliably distinguished by the CSI fingerprint, but per-individual identification is not required and is explicitly out of scope of claim 1.
- Non-occupant motion confounders. Pets, ceiling fans, blinds, and HVAC airflow can introduce CSI variance that resembles human motion. The inference module (302) employs band-limited filtering and amplitude thresholds derived from human-respiration prior (0.1–0.4 Hz) and human-body-size scattering cross-section to suppress non-human confounders. Where the confounder cannot be suppressed, the system gracefully degrades to a clock-driven baseline and surfaces a low-confidence indicator.
- Edge population. The flicker-fusion and CCT JND curves of FIG. 4 are population-mean. Individual occupants with above-average flicker sensitivity (e.g. some migraine sufferers, some autism-spectrum readers) may perceive modulation that lies safely below the population threshold. In some embodiments the operating point (FIG. 4 · 408) is calibrated per occupant via a one-time perceptual-threshold elicitation procedure.
- Newborn / infant in room. Where a sleeping infant is present, the system shifts the operating point further below the population threshold and clamps the CCT to the warmest available value, on the rationale that infant visual systems have different temporal sensitivity profiles. Infant presence is inferred from CSI patterns and is not personally identifying.
- Adversarial input. An attacker on the local LAN cannot reach occupant identity or location, because no such information leaves the dwelling boundary (FIG. 3 · 300, FIG. 8 · 802). The optional inter-room mesh traffic (FIG. 8 · 816) carries only the smoothed calm index, which is dimensionless and not personally identifying.
Part IV · Claims 6 total · 1 indep · 4 dep · 1 apparatus
1. A computer-implemented method for modulating an illumination device deployed in a dwelling as a function of a locally-inferred occupant-state index, the modulation being held strictly below a perceptual-threshold floor so as to remain peripheral to occupant attention, the method comprising:
- (a)obtaining, at a local computing device situated within said dwelling, a passive-sensor stream from at least one of (i) a Wi-Fi channel-state-information receiver (301a, per RoomMirror parent) and (ii) a portable computing device's inertial-measurement unit and microphone (301b), said passive-sensor stream containing no image data of the occupant;
- (b)computing on said local computing device an occupant-state index (302) from said passive-sensor stream;
- (c)smoothing said occupant-state index by a low-pass filter (304) with cutoff period τ_p ≥ 30 s;
- (d)mapping said smoothed index, via a configurable luminaire driver (306), to a setpoint pair comprising one or both of a brightness and a correlated color temperature of a target illumination device (308); and
- (e)continuously actuating said target illumination device with said setpoint pair such that the resulting modulation period and depth lie strictly below a perceptual-threshold curve (402) as set forth in FIG. 4, said modulation being distinct from clock-driven circadian-lighting in that the setpoint trajectory is driven by said occupant-state index rather than by time-of-day;
- (wherein)no occupant identifier, location, or raw sensor stream is transmitted off-host.
2. The method of claim 1, wherein the passive-sensor stream is a Wi-Fi channel-state-information stream per the RoomMirror parent specification, and the occupant-state index is a respiration-rate-derived calm index.
3. The method of claim 1, wherein τ_p lies in the range 30 to 300 seconds, and the modulation depth lies below 20 % of brightness and 800 K of correlated color temperature.
4. The method of claim 1, further comprising overriding said modulation to a hold state in response to a user-foreground attention event detected by said local computing device.
5. The method of claim 1, wherein the perceptual-threshold curve (402) is configured such that (i) the time-domain modulation in brightness lies below the short-term flicker indicator (Pst-LM) and the stroboscopic visibility measure (SVM) as defined by IES TM-33 (Bodington, Bierman & Narendran 2016), grounded in the dynamic-sensitivity functions of De Lange 1958 and Kelly 1961; and (ii) the correlated-color-temperature modulation depth lies below one MacAdam Standard Deviation of Color Matching (1 SDCM, ≈ 0.005 in Δu′v′) as defined by CIE 015:2018 and CIE TN 001:2014; and wherein, in some embodiments, the target illumination device is replaced by an ambient surface comprising one of a Nanoleaf-class panel, a smart blind modulating daylight transmittance, or a low-resolution e-ink ambient frame.
6. An apparatus comprising:
- (a)a local computing device coupled to said passive sensor and to a luminaire driver implementing the method of any of claims 1 – 5; and
- (b)a non-transitory memory storing the configurable mapping and perceptual-threshold curve.
| Claim | Key element | Supporting figures & numerals |
|---|---|---|
| 1(a) | passive-sensor stream (CSI or phone IMU + mic), no image data | FIG. 3 · 301a, 301b |
| 1(b) | occupant-state index computed locally | FIG. 1 · 106 · FIG. 3 · 302 |
| 1(c) | low-pass smoothing, τ_p ≥ 30 s | FIG. 2 · 208, 210 · FIG. 3 · 304 |
| 1(d) | configurable mapping to ⟨brightness, CCT⟩ setpoint pair | FIG. 1 · 102 (CCT range) · FIG. 3 · 306, 308 |
| 1(e) | continuous actuation strictly below perceptual-threshold curve; distinct from clock-driven circadian lighting | FIG. 2 · 206, 212 · FIG. 4 · 402, 406, 408 · FIG. 5 · 506 vs 520 |
| 1(w) | no off-host transmission of occupant ID, location, or raw sensor stream | FIG. 3 · 300 (LOCAL DWELLING boundary) |
| 2 | CSI as passive-sensor source; respiration-rate-derived calm index | FIG. 1 · 106 · FIG. 3 · 301a |
| 3 | τ_p ∈ [30, 300] s; depth ≤ 20 % brightness, ≤ 800 K CCT | FIG. 4 · 402, 408 · FIG. 1 · CCT range |
| 4 | user-foreground override → hold state | FIG. 3 · 306 (driver branch) · FIG. 7 · 702, 704, 706, 708 |
| 5 | perceptual bounds (flicker-fusion + CCT JND) + alternative output surfaces | FIG. 4 · 402, 406 · FIG. 6 · 602, 604, 606, 608 |
| 6 | apparatus · local compute + memory + sensor + driver | FIG. 3 · 300, 301a/b, 306, 308 · FIG. 6 · 602 |
Part V · Appendices Reference numerals · Bibliography
- 102luminaire (desk-lamp form factor)
- 104occupant (seated at desk)
- 106respiration trace
- 202deep-work band (09–15 h)
- 204wind-down band (18–22 h)
- 206lamp intensity · phase-lagged output
- 208calm-index baseline curve
- 210phase-lag annotation (≈ 30 min)
- 212minimum-depth reference (FIG. 4)
- 300local dwelling / host boundary
- 301aWi-Fi CSI receiver (RoomMirror)
- 301bphone IMU + microphone
- 302occupant-state inference
- 304calm-index LPF (τ_p ≥ 30 s)
- 306luminaire driver
- 308bulb
- 312cloud / off-host · prohibited path
- 402perceptual-threshold floor (period × depth)
- 404foreground zone · noticeable flicker
- 406peripheral envelope · operating zone
- 408operating point · 45 s, 12 %
- 410time-domain inset of operating point
- 502Hue / LIFX / Nanoleaf
- 504Apple Adaptive Brightness
- 506Circadian-lighting · f.lux
- 508Apple Focus mode
- 510Calm Automaton (CHI EA '17)
- 512Towards Calm Displays (CHI '17)
- 514Wellness lighting · BIOS / CASAMBI
- 516RoomMirror (parent · input only)
- 518Hue Sync · Apple Health hooks
- 522MoodLight (Snyder · CSCW '15)
- 524DeLight (Yu · PUC '18)
- 520LampTide (this disclosure)
- 602single smart bulb · primary embodiment
- 604multi-luminaire mesh
- 606Nanoleaf-class panel array
- 608smart blind · daylight modulator
- 702NORMAL state · modulation active
- 704HOLD state · setpoint frozen
- 706override-assert transition
- 708release + N-second cooldown
- 802dwelling boundary
- 804Room A · bedroom
- 806Room B · living
- 808a/b/cper-room inference instances
- 810a/b/cper-room luminaire instances
- 812a/b/cper-room calm-index traces
- 814Room C · kitchen
- 816inter-room mesh sync · LOCAL LAN only
- 902INIT · default operating point
- 904MODULATE · continuous loop
- 906SENSE · override-frequency monitor
- 908ADJUST · ±1% depth step
- 910update-and-loop edge
- 912per-occupant trajectory
- Calm technology · ambient information · peripheral interaction
- Weiser, M. and Brown, J. S. The Coming Age of Calm Technology. Xerox PARC, 1996.
- Mankoff, J., Dey, A. K., Hsieh, G., Kientz, J., Lederer, S. & Ames, M. Heuristic Evaluation of Ambient Displays. CHI 2003.
- Pousman, Z. & Stasko, J. A Taxonomy of Ambient Information Systems: Four Patterns of Design. AVI 2006.
- Bakker, S., van den Hoven, E. & Eggen, B. Peripheral Interaction: Characteristics and Considerations. Personal and Ubiquitous Computing 19(1), 2015.
- Hausen, D. Peripheral Interaction. Springer 2016 (peripheral output defined formally).
- Case, A. Calm Technology. O'Reilly, 2015.
- Calm Tech Institute, Calm Tech Principles & Certification Criteria. 2022–2024.
- Smart luminaires · calm-tech precedents · subtle interaction
- Cho, M. & Saakes, D. Calm Automaton: A DIY Toolkit for Ambient Displays. CHI EA 2017. [DIY toolkit · not physiological-coupled]
- Kučera, J., Scott, J., Chen, N., Olivier, P. & Hodges, S. Towards Calm Displays: Matching Ambient Illumination in Bedrooms. IMWUT / UbiComp 2017. [ambient-reflective · not state-driven]
- Pohl, H., Muresan, A. & Hornbæk, K. Charting Subtle Interaction in the HCI Literature. CHI 2019.
- Ross, P. R. & Wensveen, S. A. G. Designing Behavior in Interaction: Aesthetic Experience as a Design Mechanism (intelligent lamp case study). Int. J. Design 4(2), 2010.
- Philips Hue developer documentation; LIFX local LAN API; Nanoleaf OpenAPI — representative smart-luminaire control surfaces.
- Biosensor → ambient light · most-threatening prior art
- Snyder, J., Matthews, M., Chien, J., Chang, P. F. et al. MoodLight: Exploring Personal and Social Implications of Ambient Display of Biosensor Data. CSCW 2015. [FIG. 5 row 522 · foreground hue mapping, no slow-cycle, no perceptual-floor, GSR-tethered]
- Yu, B., Hu, J., Funk, M. & Feijs, L. DeLight: Biofeedback through Ambient Light for Stress Intervention and Relaxation Assistance. Personal and Ubiquitous Computing 22(4), 2018. [FIG. 5 row 524 · HRV biofeedback coaching loop, foreground-only]
- Howell, N., Devendorf, L., Tian, R. V. et al. Tensions of Data-Driven Reflection: A Case Study of Real-Time Emotional Biosensing. CHI 2018. [critique that supports HOLD override (claim 4) as principled, not cosmetic]
- Picard, R. W. Affective Computing. MIT Press, 1997.
- Wellness / circadian lighting
- Houser, K. W., Boyce, P. R., Zeitzer, J. M. & Herf, M. Human-centric lighting: Myth, magic or metric? Lighting Research & Technology 53(2), 2021.
- Durmus, D. Correlated color temperature: Use and limitations. Lighting Research & Technology, 2022.
- Khademagha, P., Aries, M. B. C., Rosemann, A. L. P. & van Loenen, E. J. Implementing non-image-forming effects of light in the built environment. Building and Environment 108, 2016.
- Apple Inc. Adaptive Brightness / Night Shift. macOS & iOS developer documentation.
- Visual psychophysics · perceptual-threshold prior art
- De Lange, H. Research into the dynamic nature of the human fovea-cortex systems with intermittent and modulated light. J. Opt. Soc. Am. 48(11), 777–784, 1958.
- Kelly, D. H. Visual responses to time-dependent stimuli I. Amplitude sensitivity measurements. J. Opt. Soc. Am. 51(4), 422–429, 1961.
- Wei, M., Houser, K. W., Allen, G. R. & Beers, W. W. Color Preference and Visibility of Flicker for LED Sources. LEUKOS / IES, 2014.
- CIE TN 001:2014; CIE 015 Colorimetry, 4th ed. 2018. [MacAdam SDCM / chromaticity-difference units for ΔCCT in claim 1]
- Bodington, D., Bierman, A. & Narendran, N. Stroboscopic Visibility Measure / Temporal Light Artefacts (IES TM-33). 2016. [Pst-LM / SVM standardised TLA metrics]
- Brown, E., Foulsham, T., Lee, C.-S. & Wilkins, A. Visibility of temporal light artefact from flicker at 11 kHz. Lighting Research & Technology, 2020.
- Occupant sensing · Wi-Fi CSI · phone-as-sensor
- Wang, H., Zhang, D., Ma, J. et al. Human Respiration Detection with Commodity Wi-Fi Devices. UbiComp 2016.
- Liu, J., Wang, Y., Chen, Y. et al. Tracking Vital Signs During Sleep Leveraging Off-the-shelf Wi-Fi. MobiHoc 2015.
- Adib, F., Mao, H., Kabelac, Z., Katabi, D. & Miller, R. C. Smart Homes that Monitor Breathing and Heart Rate (Vital-Radio / WiTrack). CHI 2015.
- Ma, Y., Zhou, G. & Wang, S. Wi-Fi Sensing with Channel State Information: A Survey. ACM Computing Surveys 52(3), 2019.
- Lane, N. D., Miluzzo, E., Lu, H., Peebles, D., Choudhury, T. & Campbell, A. T. A Survey of Mobile Phone Sensing. IEEE Communications Magazine 48(9), 2010.
- Sano, A. & Picard, R. W. Stress Recognition Using Wearable Sensors and Mobile Phones. ACII 2013.
- Park, C. Y. et al. K-EmoPhone: A Mobile and Wearable Dataset with In-Situ Emotion, Stress, and Attention Labels. Scientific Data, 2023.
- Smart-home privacy · on-device / edge inference
- Apthorpe, N., Reisman, D. & Feamster, N. A Smart Home is No Castle: Privacy Vulnerabilities of Encrypted IoT Traffic. arXiv 1705.06805, 2017. [motivates claim 1's no-off-host wherein-clause]
- Apthorpe, N., Shvartzshnaider, Y., Mathur, A., Reisman, D. & Feamster, N. Discovering Smart Home Internet of Things Privacy Norms Using Contextual Integrity. IMWUT 2018.
- McMahan, B., Moore, E. et al. Communication-Efficient Learning of Deep Networks from Decentralized Data (FedAvg). AISTATS 2017.
- Banbury, C. R., Reddi, V. J. et al. MLPerf Tiny Inference Benchmark. MLSys 2021. [supports feasibility of bulb-class on-device inference]
- Ren, J. et al. Tiny Machine Learning and On-Device Inference: A Survey. IEEE 2023.
- Multi-luminaire coordination · spatial ambient · override semantics
- Zigbee Alliance / Connectivity Standards Alliance, Matter 1.x Lighting Cluster specifications. 2021. [protocol anchor for FIG. 6 · 604 and FIG. 8 · 816]
- Aliakseyeu, D., Mason, J., Meerbeek, B., van Essen, H. & Offermans, S. Designing Smart Lighting for Office Environments. Eindhoven / Philips Research, 2014–2015.
- Magielse, R., Hengeveld, B. J. & Frens, J. W. Designing a light controller for a multi-user lighting environment. IASDR 2013.
- Hong, S. et al. Attentive Notifications: Minimizing Distractions of Mobile Notifications through Gaze Tracking. MobileHCI 2023. [supports sustained-gaze trigger of claim 4]
- Pielot, M., Vradi, A. & Park, S. Dismissed! A Detailed Exploration of How Mobile Phone Notifications Are Handled. MobileHCI 2018.
- Mehrotra, A., Hendley, R. & Musolesi, M. NotifyMeHere: Intelligent Notification Delivery in Multi-Device Environments. CHI 2019.
- Parent disclosure
- Kim, J. H. RoomMirror. Provisional draft v0.10 (parent application).
Part VI · Execution Version · v1.0 · Per-occupant calibration · standardised perceptual metrics
- v0.12026-04-12 · Skeleton draft. Descendant of RoomMirror specification.
- v0.22026-04-18 · Added FIG. 5 (prior-art comparison matrix); drawing-convention legend; Field of Invention block; Detailed Description; Alternative Embodiments; reference-numerals appendix; claim-figure support chart; lineage breadcrumb; accent color applied to perceptual-threshold curve, operating point, and lamp-intensity output; narrowed Claim 1 to explicitly distinguish from clock-driven circadian lighting; added Claim 5 on flicker-fusion + CCT-JND bounds and alternative output surfaces; bibliography grouped and extended with Kelly 1961, De Lange 1958, Wei 2014, CSI-respiration literature; section-divider hierarchy strengthened; Claims box bumped to /70 border + tint; Distinguished-from sidebar legibility fix.
- v0.32026-04-24 · Reworked FIG. 3 with stage groupings (SENSING/INFERENCE/SMOOTHING/ACTUATION), sensor glyphs, OR junction, filter glyph + accent LPF qualifier, and explicit no-cloud privacy marker with prohibited-path X; reworked FIG. 4 with proper log-scale x-axis, 0–30% y-axis ticks, hatched foreground region, repositioned zone labels, and time-domain inset of operating point; added FIG. 6 (hardware-embodiment taxonomy · single bulb primary + mesh/panel/blind alternatives) and FIG. 7 (override state machine NORMAL ↔ HOLD for claim 4); ref-numerals appendix gained 600s and 700s groups; claim chart cross-references updated to span FIG. 6 + FIG. 7.
- v0.42026-04-30 · Added FIG. 8 (multi-room mesh extension · three independent per-room pipelines with optional local-LAN inter-room sync 816); added Robustness & failure-modes block in Part III (sensor unavailable / multi-occupant / non-occupant motion confounders / edge-population flicker sensitivity / infant present / adversarial input); added multi-room mesh as an eighth Alternative Embodiment; reference-numerals appendix gained 800s group; Part III subtitle now lists Robustness.
- v0.52026-05-06 · Integrated focused HCI-literature scan. Added MoodLight (Snyder CSCW 2015, row 522) and DeLight (Yu PUC 2018, row 524) to FIG. 5 as the most-threatening biosensor-to-light prior art and updated the FIG. 5 caption accordingly; expanded bibliography from 5 themed groups to 8 with ~22 new entries spanning peripheral interaction (Hausen 2016 / Calm Tech Institute), wellness lighting (Houser 2021 / Durmus 2022 / Khademagha 2016), perceptual-threshold standards (CIE TN 001, IES TM-33 Pst-LM/SVM, Brown 2020), occupant sensing (Adib Vital-Radio CHI 2015, Ma CSI survey 2019, Lane phone sensing 2010, Sano & Picard ACII 2013, K-EmoPhone 2023), smart-home privacy (Apthorpe 2017/2018, FedAvg, MLPerf Tiny, TinyML survey 2023), multi-luminaire coordination (Matter Lighting Cluster, Aliakseyeu / Magielse), affective computing (Picard 1997, Howell CHI 2018), and override semantics (Hong MobileHCI 2023 sustained-gaze, Pielot 2018, NotifyMeHere); reference-numerals appendix gained 522 and 524 entries.
- v1.02026-05-12 · Added FIG. 9 (per-occupant θ calibration · flowchart 902→904→906→908 with loop 910, plus three convergence trajectories 912 over 56 days); tightened Claim 5 with standardised perceptual metrics (IES TM-33 Pst-LM/SVM, CIE 015:2018 MacAdam SDCM, CIE TN 001:2014) per the literature agent's defence recommendation; updated Distinguished-from sidebar to include MoodLight and DeLight in accent; updated About box to acknowledge MoodLight/DeLight as closest biosignal-to-light analogs and explain LampTide's perceptual-floor distinction; added Detailed Description paragraph citing Apthorpe 2017 (smart-home privacy) and MLPerf Tiny / TinyML survey to load-bear the no-off-host wherein-clause; reference-numerals appendix gained 900s group.
- v1.02026-05-18 · Portfolio-wide prior-art audit pass (no substantive changes). The audit confirmed the v1.0 prior-art posture: the headline mechanic (sub-perceptual luminaire modulation driven by an inferred occupant-state index) is not anticipated by any commercial product or US patent reviewed; the closest-on-point art (MoodLight CSCW '15 and DeLight PUC '18) was already cited and explicitly distinguished by the sub-perceptual constraint. Adjacent commercial wellness-lighting deployments (Philips/Signify rPPG biosensing licensing program, Lutron HXL Human-Centric Lighting, Apple Adaptive Brightness) are above-threshold scene controls and do not anticipate the perceptual-floor limitation. Marker bumped to v1.0 to align with the EchoCast / RoomMirror audit pass — substantively a no-op for LampTide.
This descendant filing cites the RoomMirror parent specification by reference and adds one narrow claim group (the slow temporal modulation of a luminaire below the perceptual-threshold floor as a function of a locally-inferred occupant index).