Surface EMG Based Biometric Person Authentication by a Grasshopper Optimized SVM Algorithm

Biometric techniques can provide extremely accurate and secured access to information compared to traditional person authentication methods through the use of users unique characteristic data. One of the newest and a promising physiological entity used in this field is the Surface electromyography (sEMG). However, machine learning based optimization are needed to have more accurate and precise results. In this context, the aim of this chapter is to introduce the Grasshopper optimized support vector machine to the field of person authentication based on sEMG. In this chapter, a first investigation of the introduced optimized algorithm accuracy will be conducted by the identification of 10 persons from the sEMG signals collected on their forearms via two Myoware sensors. 27 features in time and frequency domain were extracted from the collected database and graphically evaluated fo their person distinguish ability before being considered as an input to a feature selection stage. Three conventional feature selection methods which are Neighbourhood component analysis (NCA), Principal component analysis (PCA) and Relief-f where compared by their performances in combination with support vector machine (SVM) to perform the person identification. Furthermore, to insure a good performance of the person authentication, an optimization of the support vector machine by the grasshopper optimization algorithm has been implemented. A person recognition accuracy of 80% is reached by the GOA-SVM with 20 search agents and 100 iterations of GOA even without any additional feature selection process while the best performance was shown by GOA-SVM with NCA as 88% of person recognition accuracy.