Blood Pressure Waveform Decomposition for Noise-Robust Measurement

The patent applies preliminary action by pre-training the deep learning model using extensive data collection and calibration before actual measurement. The model is trained in advance on diverse pulse waveform data to learn the relationship between waveform features and blood pressure, so that during actual measurement, the system can directly use the pre-trained model without requiring time-consuming data collection or calibration procedures. This resolves the contradiction by shifting the data collection and calibration work to a preliminary phase, making the measurement process faster while maintaining accuracy.

The patent applies self-service by enabling the pre-trained deep learning model to automatically process pulse waveform data and output blood pressure measurements without requiring manual intervention for parameter setting or calibration. The model autonomously learns from the input waveform and produces measurements directly, eliminating the need for complex manual calibration procedures. This resolves the contradiction by replacing complex manual measurement process with an automated self-service system that maintains accuracy while reducing operational complexity.

The patent applies the blessing in disguise principle by converting the harmful noise in pulse waveforms into beneficial training data for the deep learning model. Instead of attempting to eliminate noise through additional data collection or calibration, the system uses the noisy waveforms as part of the training data, teaching the model to recognize and filter noise patterns while extracting useful blood pressure information. The model learns to distinguish between noise and meaningful signal features, thereby converting the previously harmful noise into a factor that improves the model's robustness and accuracy in real-world conditions.