Physiological State Recognition Model Training With GAN Signal Augmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in achieving accurate physiological state recognition due to individualized differences in physiological signals, such as electrocardiogram and electroencephalogram signals, which affect the accuracy of mental, emotional, and health state assessments.
Innovation Solution
A method is developed to train a human-factors intelligent physiological state recognition model using both collected real physiological signals and generated signals based on these real signals, employing a generator and discriminator in a generative adversarial network to suppress individualized differences, thereby enhancing the model's accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If only collected real physiological signals are used for training, then the model learns individualized differences, but the recognition accuracy decreases due to variability across subjects
Solution Approach 1:
The patent generates synthetic physiological signals that copy the statistical characteristics and patterns of real physiological signals. A generator network creates artificial ECG and EEG signals that mimic the temporal, spectral, and morphological features of actual physiological data, allowing the model to learn general patterns without overfitting to individual variations.
Solution Approach 2:
The patent transforms physiological signals into different parameter representations through signal processing operations. Features such as heart rate variability, spectral power, and temporal patterns are extracted and transformed into standardized parameters that can be compared across different subjects, reducing the impact of individualized differences while preserving state-related information.
2Measurement precision
If more collected real physiological signals are collected to improve accuracy, then the recognition accuracy improves, but the data collection time and cost increase
Solution Approach 1:
The patent performs preliminary signal processing and feature extraction during the data collection phase. Instead of collecting raw signals and processing them later, the system pre-computes relevant features such as heart rate, rhythm patterns, and spectral characteristics, reducing the amount of raw data that needs to be stored and processed while maintaining recognition accuracy.
Solution Approach 2:
The patent generates synthetic physiological signals that can be used for training without requiring additional real subject participation. The generator network produces artificial ECG and EEG signals with realistic characteristics, effectively creating unlimited training data without extending data collection time or requiring more subjects.
3Measurement precision
If a simple training approach using only real signals is used, then the training process is simple, but the model accuracy is limited due to individualized variations
Solution Approach 1:
The patent introduces a generator network as an intermediary between the training data and the recognition model. This generator acts as a mediator that transforms limited real physiological signals into expanded synthetic datasets, enabling the model to learn more robust patterns without directly processing the raw individualized variations in the original signals.
Solution Approach 2:
The patent creates a composite training dataset that combines real physiological signals with synthetic generated signals. This composite approach integrates the authenticity of real data with the diversity and quantity of generated data, creating a more robust training corpus that improves model accuracy while managing complexity through structured data fusion.
Data Source
AI summary
Provided are a method for training a human-factors intelligent physiological state recognition model which includes: obtaining a first physiological signal, a first state label of the first physiological signal, a second physiological signal corresponding to the first physiological signal, and a second state label of the second physiological signal, in which the second physiological signal is a physiological signal generated based on the first physiological signal, the first physiological signal is a collected physiological signal, the first state label indicates a physiological state corresponding to the first physiological signal, and the second state label indicates a physiological state corresponding to the second physiological signal; and training a physiological state recognition model based on the first physiological signal, the second physiological signal, the first state label, and the second state label.