Facial Expression Recognition Using VQ-HMM
Abstract:
We developed an algorithm to recognize facial expression in video data using Vector Quantization and Hidden Markov Model. A salient feature detector and tracker realized by Gabor wavelet and Multi-scale Optical flow is used to locate and track the salient features for facial expression, such as eye corners, lips, etc. The tangent vectors at the trajectories of these moving salient feature points are quantized to discrete features. Discrete Hidden Markov Model then is conducted to interpret the symbol Sequences. (This project was done with Guo, Yan; Wu, Jiankang in Kent Ridge Digital Labs, Singapore, in 2000)
Publication:
Zhang, Dongqing; Guo,
Yan; Wu, Jiankang (Singapore): "Facial Expression Recognition using VQ-HMM". Conference Proceeding of 4th World Multiconference on Systemics,
Cybernetics and Informatics, July 23-26, 2000, Orlando, Florida (SCI 2000/ISAS 2000).
Mouth/Lip contour extraction/tracking using
Parabola Models
Abstract:
We developed an algorithm to detect and track the mouth corner and lip contour, which is very useful for various applications, for example, facial expression recognition and multimodality speech recognition. The mouth corner is detected using a corner detector specially designed for mouth corner. And the lip contour is modeled as double Parabola Models. Afterwards, a regression algorithm is conducted to estimate the parameters of the Parabola models. The algorithm works in real-time, therefore it can be used in tracking purpose. However, a more sophisticated system should incorporate temporal prediction and fusion to improve the stability and robustness of tracking.
(This project was done with Guo, Yan; Wu, Jiankang in Kent Ridge
Digital Labs, Singapore, in 2000)
Publication:
Dongqing Zhang, Yan Guo, Jing Zhong, Jiankang Wu, “Real-time Mouth
Detection Using Corner Vector Detection and Parabola Models”, Proceeding of 5th
Asian Conference on Computer Vision, January 8-11, 2000 Taipei, Taiwan (ACCV 2000).