Huge number of people Suffer from amputations or neurological damages, which prevent them from interacting with their environment in a simple and normal way. According to WHO report in 2021, there are more than 30-million amputee patient all over the world and the main percentage of them are arm amputees. Most prosthetics arm that uses Brain computer interface technique (BCI) are only work in lab scale. Some of them is invasive and require dangerous medical surgery to implant electrodes in the gray matter beneath scalp or use noninvasive sensor but need huge number of electrodes to compensate the loss of EEG signals. We worked on dataset include 32 EEG electrodes according to 10-20 standardization and select only 6 electrodes that are relevant to arm movement. Our proposed system includes three main stages, preprocessing of EEG signal through making band bass, CSP filters and apply CWT. In the second stage, we trained data on pretrained network -VGG16- for feature extraction. Finally, we implement our classification model on the robotic arm. We depend on accuracy and loss to evaluate classification model and built GUI by LabVIEW to determine kinematics and dynamics calculation for robotic arm. We managed to control robotic arm that closely mimics the human arm based on EEG signal and the performance of our proposed classification algorithm is evaluated in terms of accuracy with average 90.2% which can help amputees to perform their daily lives as normal people without the need for assistance from others.