# Introduction

he first and foremost important steps towards preventing unauthorized access are user authentication. User authentication is the process of verifying identity. Biometric authentication is based on something one's physiological and behavioral characteristics. In traditional approaches, passwords and tokens were used and it can be forgotten, lost or stolen. There is also usability limitations associated with them. Recently, biometric is attracting more and more attentions. Generally, biometric is a field of technology that uses automated methods for identifying or verifying a person based on a physiological or behavioral trait. [1] These traits are always measured in different systems are the face, fingerprints, palm print, handwriting, iris, gait, and voice etc. Among them, gait recognition, as a relatively new biometric technique, aims to recognize individuals by the way they walk. The advantages of gait recognition are that it can be applied inconspicuously and it offers an ability to recognize at a distance or at low resolution.


# II.


# Related work

The Gait recognition approaches for human identification plays an important role in many applications especially in security systems. The first gait recognition approach was developed by Niyogi and Author ? : Dr. M. Pushpa Rani, Professor & Head, Dept. of Computer Science, Mother Teresa Women's University, Kodaikanal, India. E-mail : pushpa_john@yahoo.com Author ? : Mrs.D.Sasikala, Research Scholar, M.Phil(CS), Mother Teresa Women's University, Kodaikanal, India. E-mail : ashwathiram.sasi@yahoo.com Adelson on a small gait database in 1994 [2]. Consequently, the HumanID program sponsored by Defense Advanced Research Projects Agency (DARPA) [3] assists greatly in advancing automatic gait recognition. Spurred by the HumanID program, many international famous universities and research institutes, such as the University of Southampton, the Massachusetts Institute of Technology (MIT), Carnegie Mellon University (CMU), Institute of Automation Chinese Academy of Sciences, etc, have made a lot of researches on gait recognition. There are various approaches available for gait recognition which can be divided into two broad categories such as Model based and Model free approaches. Zhang et al. [4] proposed a novel two-step, model-based approach to recognize gait by employing a five-link biped locomotion human model. . Meyer et al. [5] extracted and tracked the contours of different parts of the human body. Lee et al. [6] fitted seven ellipses in the human body area, and used their locations, orientations, and aspect ratios as features to represent the gait. In general, the features used in model-based approaches are insensitive to background cluttering and noise. Model based approaches has high computational complexity and more difficult in low resolution images. However, Modelbased approaches are somewhat difficult in real environment because feature extraction process and matching is very difficult. The Model-free approaches are well suitable for real time systems because it is easy to extract the feature and computational complexity is low.


# III.


# Techniques used a) Wavelet Descriptors with ICA for Feature Extraction

In this approach, the automatic Gait recognition has been accomplished based on wavelet descriptors and independent component analysis (ICA) for the purpose of human identification at a distance. The background extraction method is applied to subtract the moving human figures accurately and to obtain binary silhouettes. The binary silhouettes are described with wavelet descriptors and convert it into ID signals to get Independent Components (ICs) of these signals using ICA. The fixing point algorithm is used for calculating the Independent Component adoption and selection. Finally using Nearest Neighbor and SVM classifiers are used for recognition. For efficient gait recognition, the information of straight lines in gait silhouettes is very important. In spatio-temporal gait representation based on the Hough transform contains more straight lines information, and is more insensitive to image noise. Hough Transform is a feature extraction technique which was proposed by Paul Hough who patented the method in 1962. This technique can be used to isolate features of a particular shape within an image [8]. Using some mathematical functions, it is possible to find imperfect instances of objects to describe the boundary curves. Since its computational complexity, Hough Transform is normally restricted to first and second order equation. Commonly, the classical Hough Transform is used for the detection of regular curves such as lines, circles, and ellipses, etc. PCA Principal component analysis (PCA) is a mathematical procedure that uses an orthogonal transformation to convert a set of observations of possibly correlated variables into a set of values of linearly uncorrelated variables called principal components. [9] Principal Component Analysis (PCA) is widely utilized to reduce the dimensionality of the data. The goal of PCA is to reduce the dimensionality of the data while retaining as much as possible of the variation present in the original dataset. PCA allows us to compute a linear transformation that maps data from a high dimensional space to a lower dimensional space. Based on these two methods the gait recognition approach was generated. In this approach, initially the preprocessing works were done as follows. First, the image sequences were aligned using some mathematical approaches. It is used to resize the various sizes of images into same size. The Fig. 4 shows several gait silhouette images after alignment. Fig. 3 : The Silhouette images after alignment Secondly the Gait cycle detection was done. Before constructing the Hough template it is very important to compute the periodicity of walking in a gait sequence. After constructing a Hough Transform the Gait Template was constructed. The Gait template was constructed using Laplacian of Gaussian methods. Using this method the edges in intensity gait images were detected. Finally the PCA technique was applied for Feature extraction. 


# Discussion on analysis

The Gait recognition using Wavelet Descriptor with ICA used two public gait databases, namely Chinese National Laboratory of Pattern Recognition (NLPR) and Chinese Xi'an University of Technology (XAUT) gait database were used to test and evaluate this approach. Here NLPR database having 20 subjects and four sequences for each views angle and have three angles, namely laterally (0 o ), obliquely (45 o ) and frontally (90 o ), XAUT database includes 10 subjects and four sequences for each views angle and have three angles, namely laterally (0 o ), obliquely (45 o ) and frontally (90 o ). The Gait recognition using Hough Transform with PCA used CASIA-A gait database for the analysis purpose [10]. All subjects walk on a straight line under normal conditions. Similar to the Wavelet Descriptors with ICA the three different view angles were [11] used to capture every subject. The database consists of 20 different persons. Each person has 4 sequences per view. The database thus includes a total of 240 (20×4×3) sequences. The length of each collected sequence varies with the pace of the walker, but the average is about 90 frames [11].  


# Conclusion

The wavelet descriptor with ICA uses SVM and Nearest Neighbour classifier for classification and recognition. The second approach uses Cumulative Match Scores (CMS) for gait recognition. Both the techniques were compared with different kinds of databases and the results were shown.
12![Fig.1: Example of gait image preprocessing[7]   ](image-2.png "Fig. 1 :Fig. 2 :")
4![Fig.4 : Gait Templates using Hough Transform](image-3.png "Fig. 4 :")
5![Fig 5 : Accuracy (%) of PCA & ICA techniques](image-4.png "Fig 5 :")
1Performance Analysis(Accuracy %)View Angles &WaveletHoughRanksDescriptorTransform withwith ICA (SVMPCA (CMS)-NN )Laterally (0 o )Rank 597.5%98.5%Rank 10100%97.5%Obliquely (45 o )Rank 592.5%100%Rank 10100%98.5%Frontally (90 o )Rank 590%100%Rank 10100%100%
			© 2012 Global Journals Inc. (US)
		
		
			

				


* 
	
		Challenge and Computational Intelligence Solutions
		
			Qinghan
		
	
	
		Technology review -Biometrics-Technology,Application
				
			2007
			2
			
		
	




* 
	
		Analyzing and Recognizing Walking Figures in XYT
		
			SNiyogi
		
		
			EAdelson
		
	
	
		IEEE Computer Society Conference on Computer Vision and Pattern Recognition
				Seattle, Wash, USA
		
			1994
			
		
	




* 
	
		The HumanID Gait Challenge Problem: Data sets, Performance, and Analysis
		
			SSarkar
		
		
			PJPhillips
		
		
			ZYLiu
		
	
	
		IEEE Trans. Pattern Anal. Mach. Intell
		
			27
			2
			
			2005
		
	




* 
	
		Human Gait Recognition at Sagittal Plane
		
			RZhang
		
		
			CVogler
		
		
			DMetaxas
		
	
	
		Image Vision Computing
		
			25
			3
			
			2007
		
	




* 
	
		Model Based Extraction of Articulated Objects in Image Sequences for Gait Analysis
		
			DMeyer
		
		
			JDenzler
		
		
			HNiemann
		
	
	
		Proceedings of International Conference on Image Processing
				International Conference on Image Processing
		
			October 1997. 1997
			
		
	




* 
	
		Gait Analysis for Recognition and Classification
		
			LLee
		
		
			WE LGrimson
		
	
	
		Proc. IEEE
				IEEE
		
	




* 
	
		Automatic Face and Gesture Recognition
		
			Int
		
		
			Conf
		
		
			2002
			
			Washington, DC
		
	




* 
	
		
			JiwenLu
		
		Gait Recognition Using Wavelet Descriptors and Independent Component Analysis Springer-Verlag
				Berlin Heidelberg
		
			2006
		
	




* 
	
		A Hough Transform Based Method for
		
			JYu
		
		
			JDuan
		
		
			KSu
		
	
	
		Gait FeatureExtraction Journal of Image and Graphics
		
			10
			10
			
			2005
		
	




* 
	
		An Angular Transform of Gait Sequences for Gait Assisted Recognition
		
			NVBoulgouris
		
		
			KNPlataniotis
		
		
			DHatzinakos
		
	
	
		Proc. IEEE Int. Conf. Image Processing
				IEEE Int. Conf. Image essingSingapore
		
			2004
			
		
	




* 
	
		
		
			Casia-A GaitDatabase
		
		
		
	
	OL/DB




* 
	
		Automatic Gait Recognition Based on Statistical Shape Analysis
		
			LWang
		
		
			TNTan
		
		
			WMHu
		
		
			HZNing
		
	
	
		IEEE Transactions on Image Processing
		
			12
			9
			
			2003
		
	




* 
	
		
			Ling-FengLiu
		
		
			WeiJia1
		
		
			Yi-HaiZhu
		
		Gait Recognition Using Hough Transform and Principal Component Analysis Springer-Verlag
				Berlin Heidelberg
		
			2009
		
	




* 
	
		
			JiwenLu
		
		
			ErhuZhang
		
		
			CuiningJing
		
		Gait Recognition Using Wavelet Descriptors and Independent Component Analysis Springer-Verlag
				Berlin Heidelberg
		
			2006