Sunday 02 February 2025
A team of researchers has made significant strides in developing an algorithm that can accurately track human hand movements using only electromyography (EMG) signals, which are electrical impulses generated by muscle activity. The algorithm, called vemg2pose, uses a combination of convolutional neural networks and long short-term memory (LSTM) recurrent neural networks to predict the movement of individual fingers and joints.
The researchers tested the algorithm on 15 participants who performed a range of hand movements, including gestures, finger pinches, and free-form movements. They found that vemg2pose was able to accurately track the movement of individual fingers and joints, even when the hands were occluded or interacting with objects.
One of the key challenges in developing an EMG-based hand-tracking algorithm is dealing with the noise and variability inherent in muscle activity signals. To address this issue, the researchers used a technique called layer normalization, which helps to reduce the impact of noise on the neural network’s predictions.
The researchers also experimented with different architectures for the LSTM component of the algorithm, including a transformer-based approach that uses attention mechanisms to weigh the importance of different input features. They found that the LSTM architecture performed slightly better than the transformer-based approach in tracking hand movements.
In addition to its accuracy, vemg2pose has several other advantages over existing EMG-based hand-tracking algorithms. For example, it can be trained using a relatively small amount of data and is computationally efficient, making it suitable for use in real-time applications such as virtual reality or gaming.
The potential applications of vemg2pose are wide-ranging and could have significant benefits for people with disabilities who struggle to control devices using traditional interfaces. For example, the algorithm could be used to develop more intuitive and accessible interfaces for people with amyotrophic lateral sclerosis (ALS) or other motor neuron diseases.
Overall, the development of vemg2pose represents an important advance in the field of human-computer interaction and has significant potential for improving the lives of people with disabilities.
Cite this article: “Accurate Hand Movement Tracking Using Electromyography Signals”, The Science Archive, 2025.
Emg, Hand-Tracking, Algorithm, Convolutional Neural Networks, Lstm, Recurrent Neural Networks, Layer Normalization, Transformer, Attention Mechanisms, Human-Computer Interaction
