Artificial Neural Networks in Classifying Hand Gestures through Electromyography
Location
Nessmith-Lane Atrium
Session Format
Poster Presentation
Research Area Topic:
Computer Science - Large Data Computing (Big Data)
Abstract
Electromyography (EMG) is the technique of collecting electrical signals from the human body for further study. These signals can be processed with Artificial Neural Networks (ANNs) for the purpose of classifying the signals into different categories such as hand gestures. The goal of the research is to create a neural network that can accurately classify many hand gestures and recognize them when the signals are fed back through the system. This would greatly benefit the fields of prostheses and gesture based computing.
An Artificial Neural Network is a computer model that approximates functions and finds correlations in data sets that would generally be out of the scope of what the human mind could comprehend. This allows obscure patterns to be recognized. The ANN is trained by sending it a data set and monitoring its results to continuously retrain the network until it reaches the desired point of accuracy. This is done by leaving some of the data points out so that the network can be tested against those points for accuracy. The larger the data set, the more accurate the network can be since it has more data to be trained off of. In this research, electromyography (EMG) is performed on the human forearm to collect the data of different hand gestures. This data is then fed into the network for training and is classified gesture by gesture. The network finds patterns in the signals for each gesture so that it can classify them and recognize the hand gestures later.
This area of research is important because it allows for complex correlations and patterns to be recognized that would otherwise be left unseen. This can lead to a deeper understanding of electrical signals in the human forearm and how they are related to different muscle movements that make up hand gestures. Also, this research could lead to different methods of ANN design that are fine tuned to analyzing EMG signals.
In a humanitarian sense, this research will make steps towards more natural control of prosthetic limbs for those people who are missing a hand. This is because much of the functionality of the human hand is controlled by the muscles in the forearm. By analyzing the EMG signals from the forearm, patterns can be recognized and used to control a prosthetic hand. However, this research can also be used for those people who are not impaired. The recognized gestures can be used for more than controlling prosthetic limbs, they can also be used for controlling any digital device. The gestures will work as commands for interfacing with a computer just like they would in the prosthetic hand. These commands could control a variety of functions on a computer and would lead to a more seamless form of human computer interaction.
Presentation Type and Release Option
Presentation (Open Access)
Start Date
4-16-2016 2:45 PM
End Date
4-16-2016 4:00 PM
Recommended Citation
Leep, Kyle, "Artificial Neural Networks in Classifying Hand Gestures through Electromyography" (2016). GS4 Georgia Southern Student Scholars Symposium. 69.
https://digitalcommons.georgiasouthern.edu/research_symposium/2016/2016/69
Artificial Neural Networks in Classifying Hand Gestures through Electromyography
Nessmith-Lane Atrium
Electromyography (EMG) is the technique of collecting electrical signals from the human body for further study. These signals can be processed with Artificial Neural Networks (ANNs) for the purpose of classifying the signals into different categories such as hand gestures. The goal of the research is to create a neural network that can accurately classify many hand gestures and recognize them when the signals are fed back through the system. This would greatly benefit the fields of prostheses and gesture based computing.
An Artificial Neural Network is a computer model that approximates functions and finds correlations in data sets that would generally be out of the scope of what the human mind could comprehend. This allows obscure patterns to be recognized. The ANN is trained by sending it a data set and monitoring its results to continuously retrain the network until it reaches the desired point of accuracy. This is done by leaving some of the data points out so that the network can be tested against those points for accuracy. The larger the data set, the more accurate the network can be since it has more data to be trained off of. In this research, electromyography (EMG) is performed on the human forearm to collect the data of different hand gestures. This data is then fed into the network for training and is classified gesture by gesture. The network finds patterns in the signals for each gesture so that it can classify them and recognize the hand gestures later.
This area of research is important because it allows for complex correlations and patterns to be recognized that would otherwise be left unseen. This can lead to a deeper understanding of electrical signals in the human forearm and how they are related to different muscle movements that make up hand gestures. Also, this research could lead to different methods of ANN design that are fine tuned to analyzing EMG signals.
In a humanitarian sense, this research will make steps towards more natural control of prosthetic limbs for those people who are missing a hand. This is because much of the functionality of the human hand is controlled by the muscles in the forearm. By analyzing the EMG signals from the forearm, patterns can be recognized and used to control a prosthetic hand. However, this research can also be used for those people who are not impaired. The recognized gestures can be used for more than controlling prosthetic limbs, they can also be used for controlling any digital device. The gestures will work as commands for interfacing with a computer just like they would in the prosthetic hand. These commands could control a variety of functions on a computer and would lead to a more seamless form of human computer interaction.
