NIR Spectroscopy Blood Glucose Prediction via Subject-Specific Model Segmentation

The patent segments the calibration process by creating multiple subject-specific prediction models instead of using a single universal calibration curve. Each subject receives personalized calibration, dividing the general population into individual segments with unique spectral characteristics, thereby resolving the contradiction between operational simplicity and measurement precision.

The patent applies local quality by tailoring the calibration parameters to each individual subject rather than using uniform calibration for all. Each subject's spectral data is processed with subject-specific parameters, making the calibration locally optimized for individual anatomical and physiological variations, thus improving accuracy without significantly complicating the overall measurement process.

The patent applies preliminary action by performing subject-specific calibration during an initial phase before actual glucose monitoring. The system collects spectral data and establishes personalized prediction models in advance, so that during routine monitoring, the pre-established models are simply applied without requiring complex real-time processing, thus achieving high accuracy while managing device complexity.

The patent uses copying by creating replicated prediction models for each subject based on their spectral characteristics. Instead of implementing a single complex universal model, the system generates simplified copies of the prediction algorithm tailored to each individual, reducing the computational complexity required during actual measurement while maintaining high precision through subject-specific parameter sets.