Feature extraction of EEG for emotion recognition using Hjorth features and higher order crossings

The Human Machine Interface (HMI) is the technology that enables direct communication between the human brain and the other external devices. Emotion recognition, thus, plays an important role in the design of HMI. Electroencephalogram (EEG) shows the internal emotional state changes of a person very effectively as compared to other traditional methods such as face recognition, gesture recognition, speech recognition, etc. Thus, EEG has gained the attention of researchers in recent years for recording brain activity in HMI design. Emotion recognition can be regarded as a pattern recognition task, hence it includes basic steps likes preprocessing, feature extraction and feature classification. Many different techniques exist for extracting important features from the EEG signal such as Principal Component Analysis (PCA), Statistical features, Discrete Wavelet Transform (DWT), Frequency domain feature extraction using Fourier Transform (FT), Short time Fourier Transform (STFT), etc. The techniques implemented in this paper are Higher Order Crossings (HOC), Discrete Wavelet Transform (DWT) and Hjorth features. A comparative study of these methods is undertaken.