Ground Truth Window Labeling for Continuous Human State Monitoring
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for detecting and monitoring human states, such as drowsiness and cognitive load, face challenges due to the dynamic and variable nature of physiological manifestations, leading to inaccurate predictions and inefficient data collection, particularly in applications requiring continuous monitoring.
Innovation Solution
Record physiological data over an extended duration to capture a comprehensive range of manifestations, segment the data into shorter analysis windows, and label each window with a ground truth label to create a robust dataset for training machine learning models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If longer recording windows are used to capture comprehensive physiological manifestations, then measurement precision is improved, but temporal resolution and response time deteriorate
Solution Approach 1:
The patent segments long recording windows into multiple shorter analysis windows. Each short window is processed independently to provide timely state predictions, while multiple predictions are aggregated to achieve accurate ground truth labeling. This segmentation resolves the contradiction by enabling both fast response (through short windows) and high accuracy (through aggregation of multiple predictions).
Solution Approach 2:
The patent performs preliminary processing by collecting and storing multiple state predictions from short windows before final aggregation. This preliminary action allows the system to prepare accurate ground truth labels in advance, ensuring both rapid response capability and high measurement precision when state changes are detected.
2Speed
If shorter analysis windows are used for continuous monitoring, then temporal resolution is improved, but measurement precision deteriorates due to insufficient capture of physiological dynamics
Solution Approach 1:
The patent merges multiple state predictions from different short analysis windows to form an aggregated ground truth label. By combining predictions from multiple short windows that collectively cover the physiological dynamics, the system achieves both high temporal resolution (through short windows) and high measurement precision (through merged predictions).
Solution Approach 2:
The patent maintains continuous monitoring by processing overlapping or consecutive short analysis windows without gaps. This continuity ensures that physiological dynamics are captured comprehensively across time, allowing short windows to achieve measurement precision equivalent to longer windows while maintaining high temporal resolution.
3Speed
If multiple short analysis windows are processed individually, then temporal resolution is improved, but model sensitivity becomes inconsistent across different manifestations
Solution Approach 1:
The patent implements feedback by aggregating multiple state predictions to determine the final ground truth label. The aggregation process provides feedback that ensures consistent sensitivity across different manifestations of human states, as the combined predictions account for variable physiological expressions that individual short windows might miss.
4Reliability
If extended duration recordings are collected to capture comprehensive manifestations, then data robustness is improved, but resource consumption and data collection time increase
Solution Approach 1:
The patent segments extended duration recordings into multiple short analysis windows that can be processed independently and in parallel. This segmentation maintains data robustness by capturing comprehensive physiological manifestations across the extended period, while reducing data collection time through efficient parallel processing of segmented data.
Solution Approach 2:
The patent creates multiple copies of short analysis windows from a single extended recording session. These copies can be used for training, validation, and testing without requiring additional subject recruitment or extended data collection, thereby improving dataset robustness while minimizing resource consumption and data collection time.
Data Source
AI summary
The present disclosure relates to methods and systems for establishing ground truth in human state detection. In one embodiment, the disclosure teaches recording physiological data over a continuous duration of at least a threshold duration that is sufficient for capturing representation of the target state in bio signals used for detection. The recorded data is segmented into shorter analysis windows, as a step toward continuous state detection, each window less than the threshold duration, and labeled with a ground truth indicative of the target state. The windows and labels are stored in non-transitory memory, for later use in training, testing, and validating one or more state prediction models. The method enhances the accuracy of state detection models by providing a more representative ground truth through prolonged recordings, which capture the variable manifestations of human states in physiological processes.