Training of kinematic and kinetic analysis of human lower extremity prosthesis movements in daily physical activities with artificial neural networks

Conventional lower extremity prostheses used today can no longer adequately meet the required needs. The reason for this is that these prostheses are inanimate and their ability to move is limited compared to a normal human lower extremity. It has been observed that people with disabilities have difficulty in basic movements such as sitting while standing, standing up while sitting and climbing stairs in daily physical activities. In addition, it has been observed that the materials used in the prostheses are problematic in terms of biomaterial compatibility. In this thesis, material analysis of human lower extremity prosthesis with static (finite element) analysis method, kinematic and kinetic analysis of these movements by examining movements in daily physical activities, and training of human lower extremity prosthesis movements in daily physical activities with artificial neural network were aimed. For this purpose, a three-dimensional solid prototype of the human lower extremity was designed with Solidworks software. While creating this design, the most suitable material selected in the prosthesis designs was selected and a static analysis of this solid model was made with the help of Solidworks Simulator. The kinetic and kinematic analysis of the statically analyzed solid model was successfully performed in the motion study in Solidworks software. Finally, the movements of the human lower extremity prosthesis in daily physical activities were trained with an artificial neural network. In this thesis, it is thought that the human lower extremity prosthesis will artificially become a living limb by using training with artificial neural network. For this purpose, the lower extremity prosthesis, which is the subject of this thesis and trained with an artificial neural network, is named SmartPro, which is the abbreviation of the name Smart Prosthesis, which means smart prosthesis. In this thesis, EMG signals are used for training with artificial neural network. The data of EMG signals was used from the public database of UCI (University Of California Irvine), named 'Daily and Sports Activities Data Set'. This data set consists of motion sensor data of 19 daily and sportive activities performed by 8 subjects (4 females, 4 males, 20-30 years old) each in their own way for 5 minutes. EMG sensor data of sitting-standing and stair climbing movements were used from this data set.