Physiological Signal Filtering Using Machine-Learned Error Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for monitoring physiological signals, such as respiratory rate, in uncontrolled environments using wearable devices like smartwatches are unreliable due to high susceptibility to motion artifacts and require specialized, expensive equipment, leading to inaccurate readings.
Innovation Solution
A system utilizing a filter machine learning model to predict errors in physiological signal extraction, segmenting data into temporal windows, and discarding windows with predicted errors above a user-tunable threshold, combined with an extraction model like a Convolutional Neural Network (CNN) to accurately extract signals from accelerometer, magnetometer, and gyroscope data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If specialized equipment is used for physiological signal monitoring, then measurement precision is improved, but device complexity and cost increase
Solution Approach 1:
The patent uses machine learning models (extraction model and filter model) to copy the functionality of specialized physiological monitoring equipment. Instead of requiring actual specialized sensors, the system uses standard smartphone sensors combined with trained ML models that replicate the measurement capabilities of medical-grade equipment, thereby achieving high measurement precision without specialized hardware
Solution Approach 2:
The patent replaces traditional mechanical/sensor-based physiological measurement systems with a computational approach using machine learning models. The extraction model and filter model substitute for physical specialized sensors, using algorithmic processing of standard sensor data to achieve physiological signal measurement that previously required specialized equipment
2Measurement precision
If data filtering is applied to improve signal accuracy, then measurement precision is improved, but computational burden increases
Solution Approach 1:
The filter model is pre-trained on labeled data containing both clean and noisy physiological signals. This preliminary training allows the model to learn filtering patterns in advance, so that during actual operation, the filtering process is computationally efficient. The heavy computational work is done during the offline training phase, not during real-time signal processing
Solution Approach 2:
The filter model automatically adapts to different signal conditions and noise patterns without requiring manual tuning or intervention. The model self-adjusts by using its trained parameters to identify and filter noise in various physiological signals, reducing the need for complex real-time computational adjustments while maintaining high signal accuracy
3Reliability
If continuous monitoring is performed, then reliability is improved, but energy consumption increases
Solution Approach 1:
The system performs physiological signal extraction and filtering at periodic intervals rather than continuously processing every sensor reading. The extraction model processes data in temporal windows at specific intervals, and the filter model applies filtering periodically, allowing the system to maintain reliable monitoring while consuming less energy compared to continuous processing of all sensor data
Solution Approach 2:
The system applies filtering selectively rather than uniformly to all data points. The filter model identifies and applies filtering only to segments of data that contain noise or artifacts, leaving clean segments unprocessed. This partial application of filtering maintains signal reliability while reducing overall computational energy consumption
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Systems and methods for filtering time-varying data for filtering and extracting a predicted physiological signal. A method including: segmenting the time-varying data into temporal windows; using a trained filter machine learning model, predicting an error for each prediction of the physiological signal for each window of time-varying data, the filter machine learning model trained using physiological signal predictions based on training time-varying data and known values of the physiological signal for the training time-varying data; discarding each window of time-varying data when the predicted error for such window is greater than a threshold; and where the window of time-varying data is not discarded, outputting at least one of the window of time-varying data and the predicted error for each prediction of the physiological signal.