International Journal of Biometrics (5 papers in press)
- Random Subspace Support Vector Machine Ensemble for Reliable Face Recognition
by Bailing Zhang
Abstract: Though many progresses have been made, face recognition is still a challenging problem which has continued to attract research interests. Most of the past efforts concentrated on the design of accurate classifiers together with appropriate feature descriptions to produce identity predictions for querying faces. There has been no hint on how reliable these predictions are. Despite the possible high accuracies from some classifiers, it may still be inapplicable in certain critical situations when incorrect predictions have serious consequences. A good example is the identification of people from a wanted list. Aiming to tackle this problem, this paper propose a highly reliable face recognition scheme by Random Subspace Support Vector Machine (SVM) ensemble. Being different to previous classifier ensembles which focus on increasing classification accuracy exclusively, the objective of the proposed SVM ensemble is to provide classification confidence and implement reject option to accommodate the situations where no decision should be made. The ensemble is created using the Random Subspace (RS) meta-learning method, together with four different feature descriptions including Local Binary Pattern (LBP), Pyramid Histogram of Oriented Gradient (PHOG), Gabor filtering and wavelet transform. The consensus degree from the ensemble's voting conforms to the confidence measure and the rejection option is accomplished accordingly when the confidence falls below a threshold. The reliable recognition scheme is empirically evaluated on several benchmark face databases including AR faces, FERET faces and Yale B faces, all of which yielded consistently high reliable results, thus demonstrating the effectiveness of the proposed approach. For example, a 99.9% accuracy was obtained with a rejection rate of 2.5% for the AR faces, which exhibit promising potentials for real-world applications.
Keywords: Reliable Face Recognition; Random Subspace; Support Vector Machine Ensemble
- A New Cow Identification System Based on Iris Analysis and Recognition
by Yue Lu, Xiaofu He, Ying Wen
Abstract: There is a growing worldwide trend to implement livestock traceability systems. This paper aims to explore how iris analysis and recognition can be utilized on cow identification to enhance cow management in its traceability system. In general, a typical cow identification system based on iris analysis includes iris imaging, iris detection, and recognition. First, the image quality of the captured sequences is assessed and a clear iris image is selected for subsequent process. Second, the inner and outer boundaries of cow iris are fitted respectively as two ellipses based on the edge images during segmentation. Then we can get the segmented cow iris on which normalization is carried out using geometric method. Finally, 2D Complex Wavelet Transform (2D-CWT) is used to extract local and global characteristics of the cow iris and the phase of the filtered cow iris is encoded as features. Experimental results indicate the effectiveness of the proposed approach.
Keywords: Animal identification, Cow iris analysis and recognition, Iris segmentation. Traceability
- Dynamic Facial Expression Analysis based on Extended Spatio-Temporal Histogram of Oriented Gradients
by Seyedehsamane Shojaeilangari, wei-Yun Yau, Eam-Khwang Teoh
Abstract: Facial expression is crucial for proper analysis of a persons face. It is an indicator of the emotion of a person and thus has attracted the attention of many researchers. In this work, a novel local spatio-temporal descriptor is proposed for motion pattern detection. The proposed feature comprises histogram of 3D gradients and the gradients variation over time to robustly describe the spatial and temporal information. It also incorporates spatio-temporal pyramid structure to handle different resolution and frame rate. To reduce the dimension of the feature, we applied Genetic Algorithm for region-based feature selection. We evaluated the performance of our proposed descriptors on facial expression recognition using the Cohn-Kanade (CK+) database. The experimental results achieved 96.10% accuracy in detecting six basic emotions. The key advantages of our proposed method are: local and dynamic processing, simple implementation, high performance, and robustness to variation of video resolution or temporal sampling rate.
Keywords: spatio-temporal histogram of oriented gradient; emotion recognition; facial expression; spatio-temporal pyramid; Genetic Algorithm
- A Persian writer identification method Using Swarm-based Feature Selection approach
by Mohsen Ebrahimi Moghaddam, Soheila Sadeghi ram
Abstract: Handwritten is one of the most famous biometrics which is processed based on image processing and pattern recognition techniques. However there are a lot of reports that have already been published on handwritten text identification methods, researchers try to improve the accuracy and speed of such methods. This paper presents an offline Persian handwritten identification method in which some new text features are extracted and best ones are selected using a swarm based approach. The essence of this feature selection method is Bees algorithm, which is a modern swarm-based meta-heuristic approach. In the proposed technique, the Adaptive Neuro-Fuzzy Inference System (ANFIS) is employed as classifier and trained by the input feature vectors. It is also compared with a Multi-Layer Perceptron (MLP) and fuzzy K-Nearest Neighbour classifiers. To test the proposed method, we have collected a handwritten Persian text dataset from 125 people who have written six sheets with five lines in each of optional Persian texts. Experimental results showed that the prediction accuracy was about 98% in average while the method training time is less than most related works. It seems this method can be extended for other languages by adjusting its parameters.
Keywords: Handwritten Identification, Feature Selection, Meta-heuristic, Beeâ€™s Algorithm, ANFIS
- Biometric verification of a subject with eye movements, with special reference to temporal variability in saccades between a subjects measurements
by Youming Zhang, Jorma Laurikkala, Martti Juhola
Abstract: We recently studied the application of saccadic eye movements, measured with video cameras, to biometric verification using subjects who receive identical stimulation. The properties of a subjects saccades may vary between measurements over the course of time, so to be useful as a means of biometric verification, the temporal variability of saccades should not distort verification results significantly. We investigated the effects of such variability by repeating the same test several times with the same groups of subjects. We found that temporal variability had only a minor effect on verification results when intervals were from a few hours to two months. Compared with the classification accuracies of approximately 90% of our earlier studies when measurements were run immediately one after another, our present verification accuracies were a few percent lower. In contrast, a long interval of approximately 16 months reduced the accuracies considerably. Our results indicate that reasonably short intervals between a subjects saccade measurements do not hinder verification based on them, while very long intervals between logins can pose a problem. Since most common electronic devices, such as computers and mobile phones, are used at frequent intervals, the analysis of saccadic eye movements seems to be viable technique for enabling biometric verification.
Keywords: Biometric verification, eye movements, saccades, classification, data analysis, temporal variability of saccades