Eye Tracking Image Interpolation for Low-Latency Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional eye tracking systems face challenges in accuracy, computational efficiency, and hardware instability, leading to screen lag and degraded user experience due to high CPU usage, tracking latency, and feature extraction errors.
Innovation Solution
An eye tracking device utilizing a transceiver, processor, and machine learning model to obtain and preprocess eye images, perform interpolation, and generate estimated and interpolated eye images, reducing CPU usage and improving tracking accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional eye tracking systems use camera-based image capture and traditional processing methods, then comprehensive eye image data can be obtained, but CPU utilization increases significantly causing screen lag
Solution Approach 1:
The patent introduces an intermediary processing layer that includes image preprocessing (cropping, normalization, noise filtering) and a machine learning model (neural network) to estimate intermediate eye images. This intermediary processing reduces the computational burden on the main CPU by handling intensive image processing tasks in a dedicated pipeline, thereby maintaining tracking accuracy while reducing screen lag and CPU utilization.
2Reliability
If hardware instability or automatic exposure adjustments occur, then complete eye images may be captured, but tracking accuracy and output quality deteriorate
Solution Approach 1:
The patent applies preliminary actions through extensive image preprocessing steps before the main processing pipeline. This includes cropping to the eye region of interest, normalizing image dimensions and intensity values, and applying noise filtering. These preliminary actions prepare the images to be more robust against hardware instability and exposure variations, ensuring consistent tracking accuracy even when input images are imperfect.
Solution Approach 2:
The system implements feedback mechanisms where the machine learning model processes sequences of eye images and uses temporal information to compensate for momentary tracking failures. The model learns from historical eye image data and provides corrected estimates that feedback into the tracking system, maintaining accuracy despite intermittent hardware issues or exposure adjustments.
3Device complexity
If traditional feature extraction methods are used, then processing can be kept simple, but tracking latency increases and accuracy decreases
Solution Approach 1:
The patent replaces traditional mechanical feature extraction methods (manual detection of pupil center, corneal reflections, iris boundaries) with a machine learning-based neural network model. This substitution allows the system to process images more efficiently by learning feature representations directly from data, reducing both computational complexity and tracking latency while improving accuracy through the model's ability to generalize from training data.
Data Source
AI summary
An eye tracking device and a method for eye tracking are provided. The method includes: obtaining an eye image set; inputting the eye image set to a machine learning model to obtain an estimated eye image; and performing eye tracking according to the estimated eye image.