The advancements in deep neural network design have led to a significant increase in the possibilities and functioning of AI-assisted medical hardware. To make use of thi s progress in the field of mobile applications or even as wearable devices, a suitable hardware-s oftware ecosystem must be identified to meet the high computation and memory demands of neural networks with minimal energy consumption. In this paper, we analyze an up-to-date heterogenous embedded platform employing a deep convolutional network for hand position recognition through electromyography signals. Our evaluation aimed to determine the optimization efforts required for the architectur e to function as a human wearable device and identify the most suitable accelerators on the given platform for this task.