Myoelectric Human Computer Interaction Using CNN-LSTM Neural Network for Dynamic Hand Gestures Recognition
Conference Paper
Overview
Research
Identity
Additional Document Info
Other
View All
Overview
abstract
Human-computer interaction(HCI) has become a trendy research field recently. Many HCI systems are based on bio-signal analysis and classification. EMG signal which is formed due to muscle activation, is used in this thesis. sEMG signals play a central role in many applications, including clinical diagnostics, control of prosthetic devices, and some human-machine interactions. These applications are commonly referred to as myoelectric control. Many factors would influence the classification in myoelectric control, and limb positions are focused on in this thesis. Two research goals in this thesis are: 1. Decease the effect of arm positions when recognizing a gestures. To tackle this issue, a CNN-LSTM neural network is introduced. Compared to Dr. Shins work, the new model is able to classify more gestures with more positions. 2. Apply the new model to a human-computer interaction system. A 7-DoF Kinova robot arm is used here. And a go and grasp task is designed to test the system. For most cases, the myoelectric control system finished the task successfully in an acceptable time longer than a joystick control. In addition, this control method is easier to master compared to a joystick control. In conclusion, this research focuses on EMG-based dynamic gestures recognition with multiple limb positions. First, the CNN-LSTM neural network, which combined the advantages of CNN and LSTM is proposed in this thesis. Then this model is used for a myoelectric control system. A 7-DoF robot arm is controlled by human gestures via the system.
name of conference
2021 IEEE International Conference on Big Data (Big Data)
