EDUCATION > Projects > Prosthetic Arm

Prosthetic Arm

EMG Controlled Smart Prosthetic Arm

Project Lead

WhatsApp Image 2021-08-14 at 9.15.43 AM.jpg

Assoc. Prof. Dr. Eng. Amir R. Ali 

Executive Deputy

122056095_3737145049669138_2391960085799623551_n_edited.jpg

Eng. Malek Mahmoud 

Project Members

DSC_8157.JPG

Abdelhameed Mubarak, B.Sc. (2019)

211531956_4058978307484446_4403922592853300642_n.jpg

Ahmed Saad, B.Sc. (2018)

Brief description for the project:

In the past there was an idea about receiving signals from a human arm to a computer device to detect what kind of gesture the human arm is practicing, which was a partial success and this escalated to the idea of improving detection accuracy. Improving the accuracy of detecting these gestures requires the knowledge of what EMG signals (Electromyography) are, and how to learn from such signals. The characterization was done by the method of cross-correlation which is an approach used as a type of comparison that gives an indication of how similar two signals are. This resulted in obtaining more accurate results for detecting a human gesture and learning from the human arm to increase its accuracy.

In this project the design and control of a mechanical exoskeleton wearable glove which can act as a sensor and also as an actuator to force a specific sensation. The glove is able to sense: finger position and force. To do the haptic interface between the exoskeleton glove and the prosthetic hand, FSR is used to know whether the prosthetic hand is holding an object or not and to give us the threshold as a feedback so we can differentiate between objects. The FSR are mounted on the prosthetic hand and we get the feedback using the exoskeleton glove. The exoskeleton glove as well as the prosthetic hand are controlled and operated using Electromyography (EMG). They can also be operated using Electroencephalography (EEG).

3232.png
amiro.png
amir'.png
32.png

Challenges of the project:

The aim of this project is to use the exoskeleton glove as a sensor as well as an actuator to force a specific sensation which can both control the hand motion, resist motion and assist motion, moreover it can be used as a physical therapy device for hand muscle rehabilitation. The glove has actuators on board such as micro servo motors to be able to simulate damping algorithms and tactile feedback with a variable gradient. This will be used to achieve the differentiation between hardness of virtual objects, such as the solidity of a rock, the springiness of a sponge, or the sudden breaking of an egg after reaching a specific force threshold. The interface is done using FSR to get the feedback threshold from the prosthetic hand to the exoskeleton glove. The glove can also be used as a grip aide to patients with partial nerve damage, or can be used in conjunction with EMG to actuate the hand of a patient with complete nerve damage of the hand. The exoskeleton glove as well as the prosthetic hand are actuated using the EMG signals coming from the muscle spiker Shield.

It is recommended to make experiments using different types of filters in order to maximize the potential of the discussed method, it is also recommended to implement the same method with different softwares like MATLAB, or languages like C++ or JAVA that may be more efficient than LabView, it is also recommended to study the efficiency of the codes and to make a run time comparison between different softwares/programming languages.


It is also recommended to make more optimizations on the device used to capture the EMG signals, as discussed before, there are many factors that affected the accuracy of the Myo band like the default hand, arm diameter and body hair. Nothing can be done to solve the decrease of accuracy due to not using the default hand, as it is a feature that is related to the body of the user and not to the device itself, however, the decrease in classification accuracy due to arm diameter and body hair are mainly because of the eight sensors not being enough to accurately capture all the essential data needed for classification, so it is recommended to increase the number of EMG sensors. One way to do it is by replacing the current EMG sensors by optical sensors which have smaller size that will make it possible to place much more than eight sensors around the human arm which will lead to more classification accuracy.