Gaze Behavior Classification for Accurate User Intent Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies struggle to accurately determine user intent based on eye movements, which are influenced by various factors such as task, state of mind, and body pose, leading to inefficiencies in user interaction with electronic devices.
Innovation Solution
A real-time gaze classification algorithm that classifies eye and head movements using video-based and retinal imaging, electrooculography, and head tracking data, independent of body pose and task, to identify gaze shifting, holding, and loss events, utilizing machine learning and physiological data for improved user intent prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If eye movement tracking is used to determine user intent, then user interaction can be enhanced, but accuracy deteriorates because eye movements are influenced by task, state of mind, and body pose
Solution Approach 1:
The patent combines multiple data sources including eye tracking data, head tracking data, and scene understanding information into a unified gaze behavior classification system. This integration allows the system to distinguish between eye movements caused by different factors (task, state of mind, body pose) and accurately determine genuine user intent, thereby resolving the accuracy problem while maintaining interaction enhancement capabilities
Solution Approach 2:
The patent introduces gaze behavior classification as an intermediary layer between raw eye tracking data and user intent determination. By classifying eye movements into discrete events (saccades, smooth pursuit, fixation, blinks) and using scene understanding as context, the system mediates the relationship between eye movements and user intent, filtering out movements unrelated to intentional interaction
2Measurement precision
If real-time gaze classification is implemented, then user intent prediction is improved, but system complexity increases due to multiple data processing requirements
Solution Approach 1:
The patent segments the gaze classification process into distinct modules: eye tracking data acquisition, head tracking data acquisition, scene understanding processing, gaze behavior classification, and user intent determination. Each module handles specific tasks independently, making the overall complex system manageable and allowing for optimized processing of each component
Solution Approach 2:
The patent performs preliminary processing of eye tracking data, head tracking data, and scene understanding information before final gaze behavior classification. By pre-processing and organizing data in advance, the system reduces the computational burden during real-time classification, maintaining accuracy while managing complexity
3Measurement precision
If multiple eye tracking technologies are integrated, then measurement accuracy is improved, but device complexity increases
Solution Approach 1:
The patent designs a unified gaze behavior classification system that can process data from multiple eye tracking technologies (video-based eye tracking, retinal/fundus imaging, electrooculography, magnetic scleral search coil) through a common framework. This multi-functional approach allows the system to accommodate different tracking methods without requiring separate processing pipelines for each technology, thereby improving measurement accuracy while controlling complexity
Data Source
AI summary
Various implementations disclosed herein include devices, systems, and methods that determine a gaze behavior state to identify gaze shifting events, gaze holding events, and loss events of a user based on physiological data. For example, an example process may include obtaining eye data associated with a gaze during a first period of time (e.g., eye position and velocity, interpupillary distance, pupil diameters, etc.). The process may further include obtaining head data associated with the gaze during the first period of time (e.g., head position and velocity). The process may further include determining a first gaze behavior state during the first period of time to identify gaze shifting events, gaze holding events, and loss events (e.g., one or more gaze and head pose characteristics may be determined, aggregated, and used to classify the user's eye movement state using machine learning techniques).