Heesung Lee

A New Gait Representation for Human Identification: Mass Vector

Gait is a new biometric aimed to recognize individuals by the way they walk. Gait recognition has recently an increasing interest from researchers due to several advantages. In this paper, we have proposed a new representation for human gait recognition which is called as mass vector. The mass vector along a given row is defined as the number of pixels with a nonzero value in a given row of the binarized silhouette of a walking person. Sequences of temporally ordered mass vector are used to represent a gait of an individual. We use the dynamic time-warping (DTW) approach for matching so that non-linear time normalization may be used to deal with the naturally-occurring changes in walking speed. Experimental results show that mass vector has a high discriminative power for gait recognition. The recognition rate is around 96.25% in a canonical viewing angle in NLPR gait database by using mass vector. Our proposed system outperforms previous works.

Neural network ensemble with probabilistic fusion and its application to gait recognition

The recognition of a person from his (or her) gait is a relatively new and promising research direction in biometrics since it is noninvasive and human friendly. Gait recognition, however, has the weakness that it is not reliable compared with other biometrics. To increase reliability, we applied a neural network ensemble with probabilistic fusion to the gait recognition problem. To improve recognition accuracy, we define belief as the posterior probability of the pattern and combine the component neural networks of the ensemble based on the belief. Experiments are performed with the NLPR and SOTON databases, and the effectiveness of the proposed method for gait recognition is demonstrated.

Multi-view gait recognition fusion methodology

This paper presents a multi-view gait recognition algorithm for identification at a distance. We make use of two well known and effective gait representations namely Motion Silhouette Image (MSI) and gait energy image (GEI). MSI and GEI inherently capture the spatiotemporal characteristics of gait. We show that the individual recognition performance of MSI and GEI can be improved by using a fusion methodology. The features for MSI and GEI images are extracted using Independent Component Analysis (ICA) which is used widely in such applications. Extreme Learning Machine (ELM) classifier is then used for classification. ELM is a multiclass classifier which offers the advantage of less time consumption and high performance. The results are fused at score level making use of fusion rules such as min and max [17] to make the algorithm robust, reliable and to improve the performance of the system. Our approach is tested on the NLPR gait database. The NLPR gait database corresponds to 20 subjects, each subject has 4 sequences and there are 3 viewing angles (0deg, 45deg and 90deg) for each person. The results on the dataset show that the fusion gives good performance for the 3 views considered in this paper.

A new genetic feature selection with neural network ensemble

A neural network ensemble is a learning paradigm in which a finite collection of neural networks is trained for the same task. Ensembles generally show better classification and generalization performance than a single neural network does. In this paper, a new feature selection method for a neural network ensemble is proposed for pattern classification. The proposed method selects an adequate feature subset for each constituent neural network of the ensemble using a genetic algorithm. Unlike the conventional feature selection method, each neural network is only allowed to have some (not all) of the considered features. The proposed method can therefore be applied to huge-scale feature classification problems. Experiments are performed with four databases to illustrate the performance of the proposed method.

Gait recognition system using decision-level fusion

Gait recognition has recently attracted increasing interest from the biometric society. In this paper, we present a gait recognition system based on the fusion of multiple gait cycles using a new gait representation. First, a gait sequence is automatically partitioned into multiple gait cycles by finding the local minima of width signal. After gait cycle partitioning, we extract a new gait feature called motion contour image (MCI) that captures the contour of the binary silhouette image of a walking individual. Finally, for human identification, the outputs of nearest neighbor classifiers are fused at a decision level based on majority voting. Our proposed system is tested on the CASIA gait dataset A. Experimental results show that the proposed system is better than or equal to previous works in terms of correct classification rate.

An efficient gait recognition with backpack removal

Gait-based human identification is a paradigm to recognize individuals using visual cues that characterize their walking motion. An important requirement for successful gait recognition is robustness to variations including different lighting conditions, poses, and walking speed. Deformation of the gait silhouette caused by objects carried by subjects also has a significant effect on the performance of gait recognition systems; a backpack is the most common of these objects. This paper proposes methods for eliminating the effect of a carried backpack for efficient gait recognition. We apply simple, recursive principal component analysis (PCA) reconstructions and error compensation to remove the backpack from the gait representation and then conduct gait recognition. Experiments performed with the CASIA database illustrate the performance of the proposed algorithm.

Gait Recognition using Sampled Point Vectors

Gait is a new biometric aimed to recognize individuals by the way they walk. Gait recognition has recently an increasing interest from researchers due to several advantages. In this paper, we have proposed a new feature vector, sampled point vector, for gait recognition based on model-free method. We choose the mean and variance of value of pixels which are sampled along to central axis of silhouette image for several frames. In contract to other system, proposed features are very simple and require low storages. Nevertheless, experimental result show sufficiently good performance. To evaluate, we use a reduced multivariate model as a classifier

An efficient genetic selection of the presentation order in simplified fuzzy ARTMAP patterns

Simplified fuzzy ARTMAP (SFAM) is used in numerous classification problems due to its high discriminant power and low training time. However, the performance of SFAM is affected by the presentation order of the training patterns. The genetic algorithm (GA) can be considered as a solution to the problem because the selection of the training pattern order is a complicated combinatorial problem in a large search space. In this paper, a new genetic ordering method for SFAM is proposed to improve the performance of the algorithm. Special genetic operators are employed in the genetic evolution. Compared to the conventional methods, the proposed SFAM demonstrates better classification performance since it can efficiently deliver the desirable properties of parents to their offspring. To demonstrate the performance of the proposed method, we perform experiments on various databases from the UCI repository.

A new probabilistic fuzzy model: Fuzzification--Maximization (FM) approach

Over the past few decades, fuzzy logic systems have been used for nonlinear modeling and approximation in many fields ranging from engineering to science. In this paper, a new fuzzy model is developed from the probabilistic and statistical point of view. The proposed model decomposes the input-output characteristics into noise-free part and probabilistic noise part and identifies them simultaneously. The noise-free model recovers the nominal input-output characteristics of the target system and the noise model gives approximation to the probabilistic nature of the added noise. To identify the two submodels simultaneously, we propose the Fuzzification-Maximization (FM). Finally, some simulations are conducted and the effectiveness of the proposed method is demonstrated through the comparison with the previous methods.

Probabilistic gait modelling and recognition

Biometric researchers have recently found considerable applicability of gait recognition in visual surveillance systems. This study proposes a probabilistic framework for gait modelling that is applied to gait recognition. The basic idea of this framework is to consider the silhouette shape as a multivariate random variable and model it in a full probabilistic framework. The Bernoulli mixture model is employed to model silhouette distribution and recursive algorithms are provided for silhouette image and sequence classification. Finally, the proposed probabilistic method is applied to benchmark databases and its validity is demonstrated through experiments.