| Crynodeb: | Emotion is mental conditions that appear spontaneously based on self
conscious effort and usually followed by physiological changes. Emotion is
especially caused by stimulation. The related information with emotional
condition by someone is communicated to all body through ECG. Therefore, it is
important to conduct a research in analysis of emotion based on the measurement
of ECG.
Data emotion was categorised according to the stimulation that had been
given to the subject by using video and music as long as ECG recorded. ICA
method analysis with FastICA algorithm could be developed to obtain emotion
feature. Feature classification was based on statistical approach from the
independent component with higher of kurtosis value. The proposed method in
classification was based on decision tree using Random Forest algorithm.
The classification result shows that the emotional recognition based on
ECG signals can be well implemented by system. The developed method
successfully classifies the emotional condition from ECG signals. The method
achieves the accuracy of 92.2% for identification of neutral emotion, 93.9% for
negative emotion and 92.1% for positive emotion. The value of ICSI is obtained
about 81.2% for neutral conditions, 88.3% for negative emotions and 85.1% for
positive emotions, it means that the system is successfully to classify individually
and effective for overall.
Keywords: Emotion, ECG, FastICA, kurtosis, Random Forest, accuracy, ICSI
|
|---|