Two-Stage EEG Gait Prediction Model for Noisy Signal Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current gait training systems face challenges in accurately predicting and rehabilitating gait patterns due to the low signal-to-noise ratio of EEG signals, individual differences, and dynamic motor performance, leading to deteriorated predictor performance over time.
Innovation Solution
A two-stage machine learning model is employed to extract gait features from EEG signals, using a combination of Temporal Convolution Blocks and Spatial Convolution Blocks, with a self-attention mechanism to mitigate the impact of time variations, and integrates real-time visual feedback for gait training and rehabilitation assessment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If EEG signals are used for gait prediction, then non-invasive brain activity monitoring is achieved, but the low signal-to-noise ratio deteriorates prediction accuracy
Solution Approach 1:
The patent divides the gait prediction task into two distinct stages: a first stage that processes EEG signals to extract initial gait features, and a second stage that refines these features to generate final gait predictions. This segmentation allows each stage to be optimized for its specific function, improving overall prediction accuracy despite the noisy input signals.
Solution Approach 2:
The patent implements a preliminary action by training the first stage model on historical gait data before actual gait prediction. This pre-training enables the model to learn robust feature extraction patterns from noisy EEG signals in advance, so that when real-time prediction is needed, the model can quickly adapt and produce accurate results even with low signal-to-noise ratio inputs.
2Reliability
If traditional prediction models are used, then implementation is simpler, but predictor performance deteriorates over time due to individual differences and dynamic motor performance
Solution Approach 1:
The patent implements a dynamic two-stage model where both stages can be updated and retrained as new gait data becomes available. The first stage learns general EEG-to-gait mappings, while the second stage adapts to individual-specific patterns over time. This dynamic structure allows the predictor to maintain stable performance while adapting to individual differences and changes in motor performance.
Solution Approach 2:
The patent incorporates feedback mechanisms where actual gait measurements are used to continuously refine and update the prediction models. The second stage model receives feedback from both the first stage predictions and actual gait data, allowing it to learn from errors and improve its accuracy over time. This feedback loop ensures the predictor remains reliable despite individual variations.
Data Source
AI summary
Disclosed herein is a system for gait training and a method of gait training. The system includes a gait prediction module configured to: using a two-stage machine learning model, extract gait features from EEG signals acquired from a subject; and determine a predicted gait based on the gait features, wherein the two-stage machine learning model includes multiple first stage blocks and multiple second stage blocks, the first stage blocks and the second stage blocks being trained based on different gait data obtained solely from the subject at different time points, wherein at least one of the second stage blocks is a feature extractor block, each of the feature extractor block corresponding to respective ones of the first stage blocks.