Drafting Patent Claims for Wearable Biosignal Technology
Patent claims for wearable biosignal technology should focus on the processing that evaluates measurement reliability, corrects noise, determines an individual's condition and controls subsequent actions, rather than on sensor readings alone.
Separate measurements from determinations
Raw values from PPG, ECG, oxygen-saturation, skin-temperature and acceleration sensors are inputs to the invention. Claims should describe how signal-quality indices or features are derived from those values, how invalid intervals are excluded or corrected, and how a condition or event is then determined.
Wear pressure, skin contact, ambient light, sweat, body motion and sensor position affect measurements. Consider time alignment across sensors, identification of motion intervals and weighting according to each signal's reliability in the context of the product's actual processing architecture.
Claims protect relationships within data processing
A typical processing sequence is:
- Acquire biosignal and motion data from one or more sensors.
- Evaluate wearing conditions and signal quality.
- Remove noisy intervals or calculate correction values.
- Determine an event by comparison with a personal baseline or dynamic threshold.
- Issue a notification, repeat a measurement, deliver content or control a device according to the result.
The independent claim should include the distinctive processing relationships. Dependent claims can specify sensor combinations, time windows, quality-index calculations, baseline updates and feedback conditions. If a sensor device performs part of the processing and a smartphone or server performs the rest, identify each actor and the data-transfer steps clearly.
Describe personalization through baseline creation and updates
The word “personalization” alone does not explain the processing. Specify the observation period used to establish a baseline, excluded data, distinctions between activity and sleep states, update intervals and constraints on abrupt changes. When population and individual baselines are both used, define their order of application and exceptions.
Medical diagnosis and wellness-oriented status information may be treated differently in regulation and product design. Describe the information the product actually produces and its subsequent actions; do not overstate diagnostic effects that have not been implemented.
Define AI features through inputs, preprocessing and outputs
The AI invention guidance from Korea's intellectual property authority (in Korean) calls for descriptions of concrete means, including training data, preprocessing and models, and the relationship between input and output data. For wearable AI inventions, the important questions are which data are processed, over which intervals, and to produce which results, rather than the model name alone.
Distinguish training from inference and describe missing-value handling, normalization, feature extraction, label definitions and use of the outputs. If training takes place on a server and inference on the device, account for model distribution, version management and device resource constraints.
Materials needed for the specification
- Sampling intervals and synchronization methods for each sensor
- Criteria for wearing-condition and signal-quality assessment
- Signal examples before and after filtering or correction
- Rules for creating and updating personal baselines and thresholds
- Comparative results for false positives, false negatives, latency and power consumption
- Allocation of functions and communication sequences among the device, app and server
Sensor datasheets alone rarely reveal the processing differences. Review algorithm descriptions, data-flow diagrams, test results and exception-handling materials together to align the claims with the product implementation.
