 This paper presents a novel approach to estimating finger joint angles from electromyogram, EMG, signals. It uses an encoder-decoder network with an attention mechanism to accurately predict the movements of multiple fingers simultaneously. The attention mechanism allows the model to learn the non-linear relationship between the EMG signals and the finger joint angles, making it more explainable than other models. Additionally, the model was tested on real-world data and found to be highly accurate. This article was authored by Hyunin Lee, Dongwook Kim, and Yongla Park.