Withthe advancement in mechanical characteristics of prosthetic hands, the need to develop a novel control strategy is crucial. Although surface electromyography (sEMG) is a functional human-machine interface methodin various commercial prostheses, it has practical limitations such as a low signal-to-noise ratio. This paper focuses on the forearmultrasound imaging method to recognizeindividual finger movement. In contrast to other published research, dedicated to only discriminating hand gestures, we presenta method to control hand prostheses by the anglesof each finger. By taking ultrasound imaging from a healthy male subject while flexing and extending his finger, and labeling them through attaching a checkerboard to the fingers, the FUMUS (Ferdowsi University UltraSound)imagesareproduced. Due to the ability of convolutional neural network to extract features, wedesignan end-to-end system for each of four deep convolutional neural networks named Visual Geometry Group Networks(VGG-16 and -19), MobileNet V1 and V2and used90% of our datasetto train the networks and validate their performance in recognizingthe label of new forearm ultrasound images. Results show approximately1 degree Mean Absolute Error (MAE)betweenthe labels of the unseen10% dataset to neural networks and the exact label of them.

Keywords