Neural Network Quantifying Mental State Intensity From Biosignals
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for determining mental states, such as cognitive load in drivers, lack reliability in quantifying intensity, relying on subjective ratings or limited data comparisons.
Innovation Solution
A computer-implemented method using an artificial neural network trained with pairs of biosignals and annotations to predict mental state intensity, processing production biosignals to quantify mental states like cognitive load, stress, or attention, through pre-processing and supervised learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If subjective ratings or limited data comparisons are used to determine mental state, then the method is simpler to implement, but the reliability and measurement precision of mental state quantification deteriorates
Solution Approach 1:
The patent replaces subjective human rating mechanisms with an automated artificial neural network system. The neural network processes biosignals (heart rate, skin conductance, eye movements) to objectively quantify mental state intensity, eliminating the need for subjective self-ratings or observer judgments while achieving higher reliability through consistent, data-driven assessment
Solution Approach 2:
The patent uses multiple biosignal measurements as copies or representations of the mental state. By collecting and analyzing multiple physiological signals (heart rate variability, skin conductance levels, eye movement patterns), the system creates a comprehensive profile that represents the mental state more reliably than any single measurement or subjective rating could provide
2Measurement precision
If more raw data is collected to improve measurement precision, then the accuracy of mental state prediction improves, but the quantity of data required increases
Solution Approach 1:
The patent extracts and selects only the most relevant features from the raw biosignal data. Through feature extraction techniques, the system identifies and uses only the critical physiological parameters (such as heart rate variability patterns, skin conductance peaks, specific eye movement characteristics) that are most indicative of mental state, filtering out redundant or less informative data to achieve high precision with reduced data quantity
Solution Approach 2:
The patent performs preliminary data processing and feature selection during the training phase. By pre-processing the biosignal data to extract meaningful features and establish baseline patterns, the system prepares optimized input representations that require less raw data during actual mental state assessment while maintaining high prediction precision
Data Source
AI summary
A computer-implemented method for quantifying a mental state, the method comprising: collecting, as a first training data subset, at least one pair of biosignals, wherein each biosignal is related to an intensity of a mental state of one or more persons; receiving, as a second training data subset, at least one annotation indicative of which biosignal of the pair of biosignals is related to a higher intensity of the mental state; training the artificial neural network on a training dataset comprising the first and second training data subsets to predict values of intensities of mental states; receiving a production dataset input comprising at least one production biosignal related to an intensity of the mental state as an input dataset; and processing the production input dataset by the artificial neural network to predict a value of an intensity of the mental state related to the production biosignal.