Gaze Prediction via Touch Interaction Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing gaze prediction methods on touch-enabled devices are power-hungry, require customized hardware, and consume significant processor cycles due to the use of image sensors and complex image processing, making them inefficient and battery-draining, especially in portable devices.
Innovation Solution
A method for low-power gaze prediction on touch-enabled devices that uses touch interactions to predict gazing areas without image sensing devices, employing a gaze prediction module that maps touch events to virtual grid cells and utilizes Bayesian inference or convolutional neural networks to estimate gaze probabilities based on historical data from usability sessions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image sensors and complex image processing are used for gaze prediction, then measurement precision is improved, but use of energy and processing resources worsen
Solution Approach 1:
The patent extracts and removes the image sensor component from the gaze prediction system, replacing it with touch event data collection. This eliminates the power-hungry image processing pipeline while maintaining gaze prediction functionality through alternative means (touch interaction analysis).
Solution Approach 2:
The patent substitutes the optical/mechanical image sensing system with an electronic software-based system that processes touch events. This replacement uses computational methods (Bayesian inference, neural networks) operating on touch data rather than optical data, significantly reducing power consumption while achieving comparable gaze prediction accuracy.
2Measurement precision
If image sensors and complex image processing are used for gaze prediction, then measurement precision is improved, but device complexity worsens
Solution Approach 1:
The patent removes the specialized image sensing hardware (customized hardware mentioned in background) and extracts only the essential data needed for gaze prediction (touch events). This simplifies the hardware requirements to standard touchscreen components already present in modern devices.
Solution Approach 2:
The patent makes the touchscreen display serve multiple functions: both as the primary user interface and as the gaze tracking sensor. The existing touchscreen infrastructure is repurposed to collect touch event data for gaze prediction, eliminating the need for separate specialized hardware and reducing overall device complexity.
3Measurement precision
If image processing is used for gaze prediction, then measurement precision is improved, but productivity worsens
Solution Approach 1:
The patent extracts and removes the computationally intensive image processing steps from the gaze prediction pipeline. By eliminating image capture, preprocessing, and analysis operations, the system achieves faster processing speeds while maintaining accuracy through direct analysis of touch event sequences.
Solution Approach 2:
The patent skips the lengthy image processing steps entirely and rushes directly to gaze prediction by analyzing touch events. This shortcut eliminates multiple processing stages (image capture, preprocessing, feature extraction, analysis) and achieves real-time or near-real-time gaze prediction with significantly reduced computational overhead.
Data Source
AI summary
Systems and method for low-power gaze prediction on touch-enabled electronic devices using touch interactions are provided. The method includes receiving an input touch interaction applied to a touch interaction area on a touchscreen display of a touch-enabled device and obtaining, by a gaze prediction module, a predicted gazing area on the touchscreen display based on the input touch interaction. A touch-enabled device has a processor, a touchscreen display and memory storing instructions to carry out the method. The gaze prediction module may utilize a model based on Bayesian inference or convolutional neural networks. Advantageously, gaze prediction is achieved without the use of image-capturing devices and with low power consumption. This can be used to develop intelligent hands-free interactions and personalized recommendation systems.