Hybrid deep learning model to classify patterns of fingers movements using single channel semg dry sensor

Many studies have been made in the field of human bioelectrical signals and their applications in Human-Computer Interface (HCI) and for medical applications like controlling prostheses for amputees, most of these studies still use the conventional ways of Machine Learning that involves hand-crafted features extraction, then train a network to learn, recognize, and classify different signals, these approaches are complex and require knowledge of the characteristics of the signal, which is time and resources expensive. This study hypothesis is to use the simplest hardware of one channel sEMG dry sensor and rely on a proposed hybrid model that uses One Dimension Convolutional Neural Network followed by Recurrent Neural Network (1D-CNN-RNN) involve to learn the features and patterns of segmented windows of the sEMG recorded signal, and to reduce the hardware cost and software computational power. The model was capable of achieving a relatively high accuracy of 98.02% of classifying 8 different fingers movements and hand gestures. Using the data from only one channel sensor is sufficient enough for the model to learn and classify the signals, and it was able to exceed the accuracy obtained by other similar recent studies when tested on their datasets.