Inspenet, August 4, 2023.
Innovative research has yielded promising results in the field of electronic device control through electrical muscle activation . This development of the control system could have applications in exoskeletons, technical wheelchairs, myoelectric bracelets and other devices.
[post_relacionado]The development of control techniques that allow anthropomorphic functions with high precision and reliability in robotic applications is more relevant than ever. Electrical signals of physiological origin, such as electromyography (EMG), electroencephalography (EEG), and electrooculography (EOG), are essential for device control in human-machine interaction.
Electromyography (EMG) and nerve conduction studies are tests used to measure the electrical activity of muscles and nerves. The latter transmit electrical signals that cause specific muscle reactions. When a muscle contracts, it generates an electrical signal that can be detected and measured.
In his thesis entitled “Real-time processing of electromyography signals through artificial neural networks”, Sebastián Suaid, who is an Electronics Engineer, focuses on the use of a digital processing tool known as an artificial neural network , which mimics the operation of the human brain. This network was trained using EMG recordings corresponding to three types of movements: twisting of the wrist, extension of the fingers of the hand, and contraction of the arm .
Innovation in handling electronic devices
The implementation of an artificial neural network is a viable methodology to improve existing systems and devices through engineering skills.
The relevance of this work lies in its ability to combine EMG signal processing with existing systems and devices, providing greater opportunities for people with reduced mobility.
Some potential applications of muscle signal processing include the monitoring of muscle activation for rehabilitation, the identification of possible pathologies, the control of exoskeletons, technical wheelchairs, prostheses and myoelectric bracelets, as well as the recognition of writing and silent speech.
The development of these devices is increasing and their usefulness is not limited only to therapeutic or rehabilitation applications, but also includes daily and domestic applications for anyone. Therefore, these new trends point towards the frequent use of myoelectric interfaces in daily life.
The myoelectric control system is possibly the simplest method for the user. It is based on the principle that every time a muscle in the body contracts or flexes, a small electrical signal (EMG) is generated due to chemical interaction in the muscle fibers.
The main advantage of myoelectric control systems lies in their ability to allow hands-free control in a non-invasive manner , which makes it possible to improve existing technologies, such as multifunctional prostheses, grip control, wheelchairs, among others.
The central core of the research focuses on the classification of EMG-based muscle activation signals. Processing of these electromyography (EMG) signals is complex due to their random nature.
Artificial neural networks, which emulate human brain processing, play a critical role in machine learning, a branch of artificial intelligence .
In this processing, supervised learning algorithms are applied that make it possible to identify EMG signal records. To characterize these signals, a mathematical tool called the Fast Fourier Transform (FFT) is used.
The results obtained in the investigation are highly promising. Through commands provided by the artificial neural network, a level of accuracy of 92% was achieved in the movements in the training set and 76% in the validation set.
The work was carried out with the support of the FaCENA Biomedical Engineering Group.
Source and photo: https://noticiasdelaciencia.com/art/47491/manejo-de-dispositivos-electronicos-a-partir-de-senales-musculares