Yanfeng Sun

Gabor-based dynamic representation for human fatigue monitoring in facial image sequences

Human fatigue is an important reason for many traffic accidents. To improve traffic safety, this paper proposes a novel Gabor-based dynamic representation for dynamics in facial image sequences to monitor human fatigue. Considering the multi-scale character of different facial behaviors, Gabor wavelets are employed to extract multi-scale and multi-orientation features for each image. Then features of the same scale are fused into a single feature according to two fusion rules to extract the local orientation information. To account for the temporal aspect of human fatigue, the fused image sequence is divided into dynamic units, and a histogram of each dynamic unit is computed and combined as dynamic features. Finally, AdaBoost algorithm is exploited to select the most discriminative features and construct a strong classifier to monitor fatigue. The proposed method was tested on a wide range of human subjects of different genders, poses and illuminations under real-life fatigue conditions. Experimental results show the validity of the proposed method, and an encouraging average correct rate is achieved.

Color face recognition based on 2DPCA

This paper presents a novel color face recognition approach based on 2DPCA. A matrix-representation model, which encodes the color information directly, is proposed to describe the color face image. The matrix-representation model defines the pixel in color face image as the basic unit, the color information of the pixel as the basic component, and then represents the color face image efficiently in the format of matrix. Based on the representation model, color-Eigenfaces are computed for feature extraction using 2DPCA. Nearest neighborhood classification approach is adopted to identify the color face samples. Experimental results on CVL and CMU PIE color face database show the good performance of the proposed color face recognition approach.

MULTISCALE DYNAMIC FEATURES BASED DRIVER FATIGUE DETECTION

Driver fatigue is a significant factor in many traffic accidents. We propose a novel approach for driver fatigue detection from facial image sequences, which is based on multiscale dynamic features...

Nonintrusive Driver Fatigue Detection

Driver fatigue is an important factor in many transportation accidents. Therefore, detecting driver fatigue is extremely important to improving transportation safety. When a driver fatigues, he will take many special visual cues on his face. In this paper, we combine visual cues from mouths and eyes systematically which characterize eye closed and yawning to infer the fatigue level of a driver. AdaBoost is used to extract the most discriminative features from the local binary pattern (LBP) features of eye areas and constructs a highly accurate classifier to get the eye visual cue. Yawning is an important evidence of driver fatigue. We detect drivers left and right mouth corners by gray projection, and extract texture features of drivers mouth corners using Gabor wavelets, and finally the mouth visual cue is extracted by using LDA to classify Gabor features. A probabilistic method based on Bayesian networks (BN) is used to fuse the two visual cues at the confidence level for fatigue detection. The proposed method has been tested on wide range of human subjects under real-life fatigue conditions of different genders, poses and illumination conditions. It yields a much more robust, reliable and accurate fatigue detection than using a single visual cue. The test data includes 4800 images from thirty peoples videos, and the average recognition rate of the proposed method is 96.79%.

Face recognition using extended Fisherface with 3D morphable model

In this paper, Fisherface is extended for face recognition from one example image per person. Fisherface is one of the most successful face recognition methods. However, Fisherface requires several training images for each face, so it cannot be applied to face recognition applications where only one example image per person is available for training. To tackle this problem, Fisherface method is extended by utilizing 3D morphable model to derive multiple images of a face from one single image. Experimental results on ORL face database and UMIST face database show that face recognition method proposed in this paper makes impressive performance improvement compared with conventional eigenface methods.

MPEG-4 compatible 3D facial animation based on morphable model

A novel 3D facial animation model is proposed in this paper to generate realistic facial animation sequence. The 3D facial animation model is constructed based on morphable model and compatible with MPEG-4. Uniform mesh resampling method is put forward to align prototypic 3D faces and facial animation principle defined in MPEG-4 is adopted to drive the 3D facial animation model. The animation model can automatically reconstruct realistic 3D face of specific person given a single image and be driven by FAP parameters to generate 3D facial animation sequence automatically. Experimental results show the 3D facial animation model proposed in this paper can effectively generate high realistic facial animation sequence automatically.

Face Recognition Assisted With 3D Face Model

This paper presents a 3D active deformable model assisted face recognition method across variations in pose and illumination. An active deformable model is built based on the prototypical 3D faces aligned by the uniform mesh resampling combined with mesh simplification based 3D face alignment algorithm. The active deformable model is first employed to reconstruct 3D face from the input 2D face image, and also exploited to estimate the rotation angles of the head according to the image. The rotation angles are used as the assisted information to tackle the variation in pose. An illumination regulating approach based on the distribution of the monocular grayscale intensity of the images is presented to tackle the variation in illumination during the face recognition. Experimental results on the CMU-PIE database show the good performance of the recognition system.

Multi-scale dynamic human fatigue detection with feature level fusion

Driver fatigue is a significant reason for many traffic accidents. We propose a novel multi-scale dynamic feature with feature level fusion for driver fatigue detection from facial image sequences. First, Gabor filters are employed to extract multi-scale and multi-orientation features from each image. Features of the same scale are then fused according to a fusion rule to produce a single feature. To account for the temporal aspect of human fatigue, the fused image sequence is divided into dynamic units, and a histogram of each dynamic unit is computed and concatenated as dynamic features. Finally AdaBoost algorithm is applied to extract the most discriminative features and construct a strong classifier for fatigue detection. The test data contains 600 image sequences from thirty people. Experimental results show the validity of the proposed approach, and the average correct rate is 99.33% which is much better than the baselines.

Dynamic Human Fatigue Detection Using Feature-Level Fusion

Driver fatigue is a significant factor in many traffic accidents. We propose a novel dynamic features using feature-level fusion for driver fatigue detection from facial image sequences. First, Gabor filters are employed to extract multi-scale and multi-orientation features from each image, which are then merged according to a fusion rule to produce a single feature. To account for the temporal aspect of human fatigue, the fused image sequence is divided into dynamic units, and a histogram of each dynamic unit is computed and concatenated as dynamic features. Finally a statistical learning algorithm is applied to extract the most discriminative features and construct a strong classifier for fatigue detection. The test data contains 600 image sequences from thirty people. Experimental results show the validity of the proposed approach, and the correct rate is much better than the baselines.

Local Linear Embedding with Morphable Model for Face Recognition

In this paper, we use local linear embedding and linear discriminant analysis for face recognition. Local linear embedding method is used to nonlinearly map high-dimensional face images to low-dimensional feature space. To recover space structure of face images, we use 3D morphable model to derive multiple images of a person from one single image. Experimental results on ORL and UMIST face database show that our method make impressive performance improvement compared with conventional Fisherface method.