Novel feature extraction technique for child emotion recognition

Human nature is unpredictable. It depends on his/her moods that relates to emotion. Emotion is a complex physiological characteristic of human being. It is highly flexible in case of children. An effective human-computer-interaction system is essential for such recognition. With fewer features effective recognition system design is still a challenge. In this paper, authors have been taken an attempt for it with child emotion. Feature extraction and its reduction are emphasized in this work. Initially Hurst parameters are used for feature extraction purpose. Next to it a novel method has been proposed as LP-VQC and is compared using Multi Layer Perceptron (MLP) network. The proposed technique has been shown as reduced feature set and with this feature set the recognition shows its effectiveness up to 80%. The comparison result shows the efficacy.