An EMG-Based, Real-Time Personal Identification Method Using an Gesture-Detection 1D Convolutional Neural Networks

With the increasing importance of personal information security, variety of biological identification methods have been put forward. However, these identification methods can be forged and falsified. Then, the identification method based on biometric with living body features such as electromyography (EMG) signal has been introduced. But, the existing studies on personal identification based on EMG signal are all based on offline processing, which is not suitable for real life scenarios. In this paper, a real-time EMG-based personal identification method, using peak detection algorithm, discrete wavelet transform, and 1D convolutional neural networks, is proposed. First, MYO armband is used to acquire the EMG signal. Then, EMG signals collected from the arm of 21 subjects are transmitted to the computer in real time through Bluetooth Module. Peak detection algorithm based on a sliding window is adopted to detect the hand-open gesture in real time. Once the gesture is detected, discrete wavelet transform is triggered to extract the features of the detected gesture. Finally, these extracted one-dimensional features are fed to 1D convolutional neural network to identify subjects. The result shows that the identification accuracy for 21 subjects under the hand-open gesture could achieve 98.41 % and the processing time between gesture event and identification is 37ms.