EEG Eye Tracking via Neural Network Signal Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for determining eye position and detecting eye gestures using EEG signals require complex electrode arrangements and are prone to interference, making them inaccurate and unreliable, especially with a small number of electrodes.
Innovation Solution
A method using a data-based eye movement model with a neural network comprising recurrent layers (LSTM, GRU, or transformer) and neuron layers to evaluate EEG signals from a small number of contact electrodes, processing the signals to remove interference and specifically evaluating signal components from eye muscles.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a small number of EEG electrodes are used, then device complexity is reduced, but measurement precision and reliability of eye tracking deteriorate
Solution Approach 1:
The patent transforms the EEG signal parameters by applying signal processing techniques (filtering, normalization) and feeding them into a neural network model that learns optimal parameter representations for eye position prediction, enabling accurate eye tracking with fewer electrodes
Solution Approach 2:
The patent replaces complex mechanical electrode arrangements with a data-driven neural network model that computationally compensates for the reduced sensor count, substituting physical complexity with algorithmic processing
2Ease of operation
If EEG signals are used for eye tracking, then non-invasive measurement is achieved, but signal reliability deteriorates due to interference
Solution Approach 1:
The patent extracts relevant eye movement components from the noisy EEG signals by training the neural network to identify and isolate the specific signal patterns associated with ocular activity, separating them from interfering brain signals and other noise
Solution Approach 2:
The patent converts the noisy, interference-ridden EEG signals into useful eye tracking information by using the neural network to learn from the raw signals including their noise components, transforming what was previously harmful interference into trainability data that improves robustness
3Measurement precision
If complex calculation steps are used to determine eye position, then measurement precision improves, but device complexity increases
Solution Approach 1:
The patent replaces complex mathematical calculation steps with a trained neural network model that performs eye position estimation through learned patterns, substituting algorithmic complexity with a pre-trained computational system that processes signals more efficiently
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables reliable eye tracking with a small number of electrodes, improving accuracy and tolerance to noisy signals, and allowing for effective detection of eye positions and gestures, even in changing lighting conditions or when the user blinks.
Implementation Method 1
acquiring a profile of at least one EEG signal over time by means of a contact electrode in order to obtain a signal time series of the at least one EEG signal
Data Source
AI summary
A method is disclosed for determining an eye position and/or eye gesture, in particular for use in controlling a function of an application system, in particular a wearable. The method includes (i) recording a profile of at least one EEG signal over time by way of a contact electrode in order to obtain a signal time series of the at least one EEG signal, (ii) providing input data sets from a signal block which arises from the signal time series, for successive time windows, and (iii) evaluating the input data sets in a data-based eye movement model in order to obtain the eye position or eye gesture, the eye movement model being formed using a neural network with one or more recurrent layers, in particular a respective LSTM layer or GRU layer or a transformer layer, and one or more neuron layers.