Implementing Hand Gesture Recognition Using EMG on the Zynq Circuit

This article proposes a hardware design of hand gesture recognition and its implementation on the Zynq platform (XC7Z020) of Xilinx. This proposed system is aimed to be embedded on the robotic prosthesis to improve the daily livings upper-limb amputees. Specifically, we design an architecture to identify hand movements using the Vivado HLS tool by exploiting the electromyography signal. The proposed architecture consists of creating two necessary intellectual properties (IPs) on hardware designed, tested, and validated against the software implementation. The first one performs feature extraction from the electromagnography (EMG) signal, and the second one implements the classification using the
${k}$
-nearest neighbor (k-NN) algorithm. Our framework process EMG signals acquired using an myo sensor with eight channels. The optimization of our design using pipeline directive achieves speed improvements of
$5\times $
and
$2.15\times $
for the feature extraction and predict IPs, respectively, with moderate area resource consumption and the same performance as software implementation.