Towards the Appropriate Feature Extraction and Event Classification Methods for NIRS Data

This work aims to find the appropriate feature extraction as well as classification method for different event classifications from NIRS data. NIRS is used to measure the oxygenated and deoxygenated hemoglobin concentration (HbO & HbR) in the super-facial layers of the human cortex. Using NIRS data, BCI can be achieved and to achieve BCI feature extraction and classifications are two major steps. Many features have been proved to be unique and good enough to use in all brain related medical applications. However, different statistical features extracted by principal component analysis (PCA), Non-linear Principal Component Analysis (NLPCA), and Independent Component Analysis (ICA) show different discriminative ability for event classifications. In addition, there exists a number of classification methods like Artificial Neural Network (ANN), Support Vector Machine (SVM), k-nearest Neighbor Algorithm (kNN) etc. those can be applicable to classify the NIRS data. In this study, we have investigated different statistical feature extraction and classification methods using NIRS data of left and right-hand movement from University of Canberra, Australia. Eventually, the goal is to explore the most suitable feature extraction and classification methods for NIRS data.