Facial Emotion Recognition Based on Facial Motion Stream Generated by Kinect
Nattawat Chanthaphan, Keiichi Uchimura, Takami Satonaka, Tsuyoshi Makioka · 2015
Nowadays, the human facial emotion recognitionhas been used in wide range of applications that is directlyinvolved in a human life. Due to the fragility of humans, theperformance in these applications has to be improved. In this paper, we describe the novel approach to extractthe facial feature from moving pictures. We introduce thefacial movement stream, which is derived from the distancemeasurement between each pair of the coordinates locatedon human facial wireframe flowing through each frame ofthe movement. We have proposed the Facial EmotionRecognition Based on Facial Motion Stream generated byKinect employing two kinds of facial features. The first onewas just a simple distance value of each pair-wisecoordinates packed into 153-dimensional feature vector perframe. The second one was derived from the first one basedon Structured Streaming Skeleton approach and it became765-dimensional feature vector per frame. We have presented the method to construct the datasetby ourselves since there was no dataset available for ourapproach. The facial movements of five people were collectedin the experiment. The result shows that the averageaccuracy of SSS feature outperformed the simple distancefeature using K-Nearest Neighbors by 10% and that usingSupport Vector Machine by 26%.