ECG signal processing
R-peak detection, heart-rate estimates, rhythm regularity and exploratory waveform features.
Technology / (AI) TinyML & AI
(AI) TinyML at the edge. Evidence-aware AI at the core.
We are investigating whether lightweight embedded models can help extract useful evidence from physiological measurements while explicit quality checks and safety rules govern decisions.
PPG signal processing
Optical pulse extraction, waveform timing, and signal-quality checks.
EDA / GSR
Experimental skin conductance measurements as supporting context.
Motion + temperature
Artifact interpretation and physiological context, not standalone cardiac diagnoses.
On-device (AI) TinyML
Resource-conscious experimental inference; advisory outputs may be unavailable or insufficient.
Personalized analysis
Research into baseline deviation, trends, persistence and corroboration.
AI decision approach
Confidence matters as much as the signal.
Acquire
Check quality
Extract features
Assess evidence
Recheck / respond
Blue · Insufficient evidence
Do not force a risk classification when contact, signal quality, or corroboration is inadequate.
Green / Yellow / Red
Proposed visual states for experimental low, caution and elevated concern levels; thresholds require validation.
Cardio AI Ring is an experimental research-stage concept. It is not clinically validated to diagnose or predict cardiac conditions, is not FDA-cleared, and must not be relied on for medical decisions or emergency monitoring.
AI / Decision intelligence
More than heart-rate tracking.
The Cardio AI Ring research architecture is designed to assess the quality and meaning of multiple physiological measurements before suggesting a response. (AI) TinyML contributes advisory pattern-recognition evidence; a separate, rule-based decision engine retains final authority.
01 / Sense
ECG, PPG & context
ECG, PPG & context
02 / Qualify
Noise & contact
Noise & contact
03 / Analyze
Features & (AI) TinyML
Features & (AI) TinyML
04 / Corroborate
Time & signals
Time & signals
05 / Respond
Or withhold
Or withhold
Know the baseline
Investigate meaningful changes relative to an individual's usual patterns, rather than using only generic thresholds.
Check before concluding
Compare ECG and PPG findings and evaluate motion, signal quality and persistence before escalating concern.
Compute close to the sensor
Explore lightweight (AI) TinyML inference on embedded hardware, with potential latency and power benefits.
Insufficient evidence is an answer
If measurements are unreliable or contradictory, the proposed system can withhold classification rather than guessing.
Recheck when necessary
Investigate follow-up measurements and persistence checks when initial evidence is ambiguous.
Rules remain in control
Machine-learning outputs are advisory. Explicit decision logic and safety constraints govern any proposed alert.
Illustrative scenario
When optical heart-rate readings look unusually high
The system would first consider movement and optical contact, seek consistency with ECG where available, and evaluate whether the finding persists. If evidence remains insufficient, it could display a blue insufficient-evidence state rather than issue a misleading cardiac risk classification.
These are proposed research features and potential engineering advantages, not measured reductions in false alarms or clinically validated detection performance.
Illustrative sensor signals
Two different ways to observe cardiac activity.
Multimodal sensing / Research architecture
Five sensing categories. One intelligent platform.
The Cardio AI Ring research roadmap combines electrical and optical cardiac signals with physiological and motion context. Individual sensor modules are being evaluated; full integration and clinical performance remain unvalidated.
Individual sensor illustrations are conceptual and do not depict finalized production hardware or exact component packages.
♥
01 / ECG
Electrical cardiac activity
Investigates heart rhythm and waveform features through an electrical sensing channel.
◉
02 / PPG
Optical pulse sensing
Observes blood-volume pulse patterns and offers complementary evidence to ECG.
≈
03 / EDA · GSR
Skin conductance
Explores electrodermal changes associated with autonomic arousal as contextual evidence.
⌁
04 / MOTION
Movement and activity
Accelerometer and gyroscope signals help flag movement-related measurement artifacts.
°
05 / TEMPERATURE
Skin temperature trends
Tracks temperature changes that may support longitudinal baseline interpretation.
ECG + PPG + EDA/GSR + Motion + Temperature
↓
Signal quality → Features → (AI) TinyML (advisory) → Rule-based decision engine
↓
Proposed Bluetooth smartphone app · Normal / Warning / Emergency · Insufficient evidence
Conceptual architecture only. Sensor capabilities, integrated operation, AI classification, and mobile alerts have not been clinically validated. The system is not intended for emergency monitoring.
Proposed cardiac status indicators
Clear indications, backed by evidence.
The Cardio AI Ring research design explores an easy-to-understand status system. Each proposed state reflects the decision engine's interpretation of available signals—not a medical diagnosis.
Normal
Green · No current concern identified
Available, sufficiently reliable measurements do not meet the proposed thresholds for a concerning pattern. This does not guarantee that the wearer is healthy or rule out a cardiac condition.
Warning
Yellow · Potential concern
An unusual, sustained, or corroborated physiological pattern may warrant attention. The proposed app could prompt repeat measurement or recommend contacting a healthcare professional.
Emergency
Red · High concern
A proposed high-priority alert for patterns that may warrant urgent evaluation. This indication cannot confirm a medical emergency, and the experimental ring must never be relied on to detect or rule one out.
Blue · Insufficient evidence
When contact is poor, movement interferes, or signals disagree, the system may withhold a classification rather than make an unreliable judgment.
These are conceptual, unvalidated alert states. The Cardio AI Ring is not FDA-cleared and is not intended for diagnosis, treatment, or emergency monitoring. If someone has symptoms of a medical emergency, they should seek emergency assistance regardless of any device status.
Proposed on-ring alerts & mobile notifications
See it. Feel it. Receive it on your phone.
The Cardio AI Ring concept combines visible LED indicators, an optional vibration motor, and Bluetooth communication with a companion smartphone app to make proposed status changes easier to notice.
Green LED · Normal
Indicates that available measurements do not currently trigger the proposed concern rules. Not a guarantee of cardiac health.
Yellow LED · Warning
Indicates a potentially unusual or persistent pattern. The app could show an explanation and prompt a recheck or medical consultation.
Red LED · High concern
Proposed urgent-priority indication accompanied by a noticeable vibration pattern and smartphone alert, if supported and enabled. This is not a validated emergency detector.
Blue LED · Insufficient evidence
Indicates that contact, signal quality, or corroboration is inadequate for a reliable assessment. The app could ask the wearer to adjust the ring or repeat measurement.
Ring LED + vibrationLocal visual and tactile feedback
Bluetooth Low EnergyTransmit status to a paired smartphone
App notificationDisplay the alert and supporting measurements
Optional SMSPhone-based message to a chosen contact, subject to permissions and connectivity
Proposed emergency workflow: a high-concern event could trigger a red LED and vibration on the ring, followed by a notification in the connected phone app. With the user's explicit opt-in and appropriate phone messaging capabilities, the app could offer an SMS notification to a designated contact. The ring itself does not send SMS directly.
All LED, vibration, Bluetooth alert, and SMS behaviors are planned concepts, not verified product functions. SMS delivery requires compatible smartphone functionality, permissions, and network service; messages are not guaranteed. The experimental device is not intended for emergency monitoring or as a substitute for calling emergency services.