Electromyography Gesture Recognition Using Sliding Windows for Noise Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional electromyography-based gesture recognition technologies face challenges in accurately and timely recognizing gestures due to the inconvenience and delay of wearable sensors, leading to poor robustness and interference in noise recognition.
Innovation Solution
A method and apparatus that utilize a sliding window approach to divide electromyography signals, denoise them, and recognize gesture information using a neural network model with weighted probabilities to improve accuracy and robustness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If wearable sensors are used for gesture recognition, then gesture information can be obtained, but recognition accuracy is reduced due to noise interference and delay
Solution Approach 1:
The patent divides the electromyography signal into multiple signal segments using a sliding window approach. Each segment is processed independently through the neural network model, and the results are aggregated. This segmentation allows the system to capture gesture information from multiple time windows, improving recognition accuracy while reducing the impact of noise in any single segment.
Solution Approach 2:
The patent introduces a denoising autoencoder as an intermediary component between the raw electromyography signal and the gesture recognition model. This intermediary structure learns to distinguish between noise and useful signal features during training, effectively filtering out noise interference before the signal is used for gesture recognition, thereby improving reliability.
2Productivity
If conventional gesture recognition methods are used, then gesture information can be recognized, but timeliness is reduced due to processing delays
Solution Approach 1:
The patent performs preliminary denoising and feature extraction using the denoising autoencoder before gesture recognition. By pre-processing the electromyography signals to remove noise and extract meaningful features in advance, the system reduces the computational burden during actual gesture recognition, thereby improving timeliness without sacrificing accuracy.
Solution Approach 2:
The patent replaces traditional signal processing methods with a deep learning-based neural network model. This substitution enables more efficient processing of electromyography signals, as the neural network can learn complex patterns directly from the data without requiring multiple sequential processing steps, thus reducing recognition delay and improving productivity.
3Reliability
If simple gesture recognition models are used, then processing is faster, but robustness is reduced in different usage environments
Solution Approach 1:
The patent designs a universal neural network model that can handle multiple gesture types and adapt to different usage environments. The model is trained on diverse electromyography data from various contexts, enabling it to generalize well across different scenarios. This universal approach provides robustness without requiring separate specialized models for each environment, balancing reliability with manageable complexity.
Solution Approach 2:
The patent employs a dynamic sliding window mechanism that adapts to different gesture durations and speeds. The window size and step parameters can be adjusted based on the specific application requirements, allowing the system to maintain robustness across varying usage conditions. This dynamic approach enables the model to capture relevant features regardless of gesture tempo or environment, enhancing reliability without excessive complexity.
Data Source
AI summary
This application provides a gesture information processing method and apparatus, an electronic device, and a storage medium. The method includes: acquiring an electromyography signal sample generated by an electromyography signal collection target object in connection with performing multiple gestures; dividing the electromyography signal sample through a sliding window having a fixed window value and a fixed stride into different electromyography signals of the target object; and applying the different electromyography signals to a first neural network model to determine gesture information matching the multiple gestures performed by the target object.