Emotion detection based on EEG signals by applying signal processing and classification techniques

Pattern recognition has applied in many areas in order to find unknown patterns, signs, and figures through pattern space. In our day, it was used not only in engineering areas but also in industry, automotive, physics, astronomy, biology, healthcare services and security systems to specify the patterns. Pattern recognition consists of two parts which are feature extraction and classification. Feature extraction is the vital part of the recognition since it affects the performance and the accuracy of the classification process. In this thesis, EEG based emotions were classified with pattern recognition techniques to determine the positive-negative emotions. To collect the key features, wavelet decomposition, entropy values, time-frequency analysis and statistical methods were applied. In the last part of the process, three different classifier algorithms –artifical neural network, support vector machines and k nearest neighbor were used discriminate emotions depending upon collected key features. EEG signals were collected from well-known and publicly available dataset from DEAP. Besides, the proposed method was used on 25 different subjects to observe the performance and accuracy of the discrimination process. The accuracy ratio designated between 85,0%-90,0% with the recommended technique.