Physiological State Detection Using Change-Point Voting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional systems fail to account for individual differences among human subjects and require extensive data for training, making it challenging to reliably identify abrupt physiological changes such as emotions, stress, and drowsiness using deep learning algorithms.
Innovation Solution
An unsupervised machine-learning approach that uses change point detection, voting processes, and clustering to identify critical change points in physiological signals, allowing for accurate segmentation and probabilistic analysis to detect anomalies without requiring years of data, and includes a feedback mechanism to adjust algorithm thresholds based on ground-truth data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deep learning algorithms are used for anomaly detection, then measurement precision improves, but loss of time increases due to extensive data training requirements
Solution Approach 1:
The patent segments the anomaly detection process into multiple independent change point detection algorithms that operate in parallel on different physiological signal features. Each algorithm detects change points independently, and their results are combined through voting. This segmentation eliminates the need for extensive training data while maintaining detection accuracy.
Solution Approach 2:
The patent introduces a voting mechanism as an intermediary that aggregates results from multiple simple change point detection algorithms. This voting process mediates between individual algorithm outputs to produce the final anomaly detection result, achieving high precision without requiring deep learning training data.
2Reliability
If conventional anomaly detection systems are used, then device complexity is reduced, but reliability worsens due to inability to account for individual differences
Solution Approach 1:
The patent implements dynamic adaptation by continuously updating baseline physiological profiles for each individual subject over time. The system learns each subject's unique baseline state and variations, allowing reliable anomaly detection that accounts for individual differences without requiring complex pre-training on population data.
Solution Approach 2:
The patent creates a universal anomaly detection framework that works across different individuals and physiological conditions using the same multi-algorithm voting approach. The system is universally applicable to various subjects while adapting to their individual characteristics through continuous baseline learning.
Data Source
AI summary
Systems and methods for identifying physiological states of human subjects are disclosed herein. In one embodiment, a system receives, from one or more sensors, physiological signals of a human subject. The system processes the physiological signals to extract feature signals as time series. The system identifies change points in the feature signals. The system identifies critical change points in the feature signals by applying a voting process to the change points in the feature signals. The system partitions the feature signals into segments based on the critical change points. The system detects a predetermined physiological state (an anomalous physiological state) of the human subject by applying clustering and probabilistic analysis to the segments. In response to detecting the predetermined physiological state, the system automatically takes an action to assist the human subject.