Eye Movement HMM Co-Clustering for Spatial-Temporal Patterns
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing eye movement data analysis methods fail to account for temporal and spatial information of eye movements, leading to inaccurate predictions of human behavior and cognitive processes, particularly in tasks involving stimuli with different layouts, and do not effectively measure individual differences in eye movement patterns.
Innovation Solution
The use of Eye Movement analysis with Hidden Markov Model (EMHMM) with co-clustering and Eye Movement analysis with Switching Hidden Markov Model (EMSHMM) to model eye movements, taking into account both spatial and temporal information, and allowing for the detection of common patterns across different stimuli layouts, including tasks like website viewing and visual search.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional classification methods are used to analyze eye movement data, then the ability to distinguish between different observers is improved, but the understanding of overall eye movement patterns and their association with cognitive style is lost
Solution Approach 1:
The patent segments eye movement data into two key dimensions: spatial patterns (where observers look) and temporal patterns (when and in what sequence they look). By analyzing these dimensions separately and then integrating them through HMM clustering, the method preserves both the distinguishing features and the overall pattern information that traditional classification loses.
Solution Approach 2:
The patent introduces temporal dimension analysis to complement spatial analysis. By modeling eye movements as sequences of fixations over time and using HMMs to capture temporal dependencies, the method adds a time dimension to the traditional spatial-only analysis, thereby recovering overall pattern information while maintaining observer distinction capability.
2Measurement precision
If HMMs are used to model individual eye movement patterns, then the capture of temporal and spatial information is improved, but the computational complexity and data processing requirements increase
Solution Approach 1:
The patent uses HMMs with a unified structure that serves multiple functions: modeling spatial fixation patterns, capturing temporal sequence dependencies, and enabling cluster analysis of multiple observers simultaneously. This multi-functionality reduces the need for separate analysis pipelines and simplifies the overall computational workflow despite the sophisticated underlying model.
Solution Approach 2:
The patent employs variational Bayesian inference to automatically determine the optimal number of hidden states (ROIs) and their parameters from data. By using probabilistic parameter estimation and automatic model selection, the method avoids manual tuning and reduces computational burden compared to fixed-parameter approaches, while still capturing complex temporal-spatial patterns.
3Adaptability or versatility
If deep learning methods are used for eye movement analysis, then the modeling capability is improved, but the requirement for large amounts of training data increases
Solution Approach 1:
The patent replaces data-hungry deep learning neural networks with probabilistic HMMs that rely on statistical principles and psychological theory. The HMM framework uses generative probabilistic models that can capture complex patterns with limited data by leveraging the Markov assumption and variational Bayesian inference, substituting the mechanical neural network approach with a statistically rigorous alternative that requires far fewer training samples.
Solution Approach 2:
The variational Bayesian HMM approach is self-organizing and automatically determines the optimal number of ROIs and their parameters from the data itself without requiring manual specification or large datasets. The model adapts to the data structure autonomously, extracting meaningful patterns and organizing observers into clusters based on their eye movement characteristics, thereby achieving sophisticated modeling with minimal data requirements.
4Ease of operation
If predefined ROIs are used for eye movement analysis, then the analysis process is simplified, but the ability to account for individual differences and temporal patterns is reduced
Solution Approach 1:
The patent transitions from static predefined ROIs to dynamic, data-driven ROI identification. The HMM model automatically learns the optimal set of ROIs and their temporal transition patterns from individual observer data, allowing the analysis to adapt to each observer's unique visual scanning behavior. This dynamic approach preserves simplicity in the analysis framework while dramatically improving the capture of individual differences and temporal patterns.
Data Source
AI summary
Provided are an Eye Movement analysis with Hidden Markov Model (EMHMM) with co-clustering, an Eye Movement analysis with Switching Hidden Markov Models (EMSHMM) to analyze eye movement data in cognitive tasks involving stimuli with different feature layouts and cognitive state changes, a switching hidden Markov model (SHMM) to capture a participant's cognitive CN state transitions during the task and an EMSHMM to assess preference decision-making tasks with two or more cognitive states. The EMSHMM provides quantitative measures of individual differences in cognitive behavior/style, making a significant impact on the use of eye tracking to study cognitive behavior across disciplines.