No abstract

(PCA) [8]. Tan et al. [9] proposed a local SOM approach called SOM-face method while Wright et al. [10] used Sparse representationbased methods (SRC) to reconstruct the test images by linear combination of training images. Later, Qiao et al. [11] introduced Sparsity Preserving Discriminant Analysis (SPDA) which is based on graph-based semi-supervised dimensionality reduction approach. Jiwen et al. [12] proposed Discriminative Multimanifold Analysis (DMMA) that treats the face recognition problem as manifold matching and maximizes the manifold margins of dissimilar persons by learning multiple DMMA feature spaces. Recently, Yin et al. [13] proposed a method called Double Linear Regression (DLR) whose objectives are to find best discriminating subspace and preserve the sparse representation structure. Some other popular appearance-based methods are Local Binary Patterns (LBP) [7], Uniform Pursuit (UP) [14], and Partial Distance Measure (PDM) [15]. Nevertheless, Gabor Wavelet (GW) [16-18] has been one of the prominent and successful appearance-based feature representation whose effectiveness is attributed to its biological relevance [6, 16-19]. GW uses kernels similar to receptive field on cortical cells with inherent spatial locality and is orientation selective, thus optimally localized in both space and frequency domains. Some popular implementations based on GW are Elastic Bunch Graph Matching (EBGM) [20], Gabor Fisher Classifier (GFC) [21], Local Gabor Binary Pattern Histogram Sequence (LGBPHS) [22], and Histogram of Gabor Phase Patterns (HGPP) [23]. Su et al. [24] proposed weighted fusion of Local Gabor Feature Vector (LGFV) and global Fourier transform called Hierarchical Ensemble Classifier (HEC). Besides, Jie et al. [25] proposed Local Matching Gabor (LMG) where ensembles of Borda count classifier were used to classify the Gabor features independently. Later, several improvements to LMG have been proposed over

This study investigated the skeleton points in the three dimensional (3D) space using Kinect sensor to be utilised as recognition of human based on their walking gait in frontal view. Firstly, walking gait of 30 subjects is captured using Kinect sensor. Next, all the twenty skeleton joints in the 3D space within one full gait cycle of each subjects are extracted as features. Further, the significant features amongst the skeleton joints are extracted using Orthogonal Least Square (OLS) technique followed by Locally Linear Embedded (LLE) as feature selection. The effectiveness of OLS as feature extraction along with LLE as feature selection is evaluated using Support Vector Machine as the machine classifier. Classification results attained for human gait recognition are 98.67% and 97% based on RBF and polynomial kernel respectively This proven that the significant gait features identified using OLS and LLE techniques are indeed suitable for gait recognition purpose.

Automated security systems are a useful addition for today's home where safety is essential. Vision-based security systems have the greater advantages over the traditional security systems such as using the lock, observing by the security guard, using alarm signal etc. This paper proposed an integrated dual-level vision-based home security system, which consists of two subsystems -a) movement detection and b) hand verification system. The primary movement detection technique is used to detect any movement first and the system verifies the authorized person for any secured place. It will check the threshold value where if the threshold level exceeds and the verification flag is off, the alarm will be triggered. Otherwise, if verification flag is on, it means the person is authorized and movement detection will be turned off for this person. On an event of a failure in the primary system, the secondary hand geometry verification module can act as a reliable backup to detect authorized person in a restricted area. Several experiment results have shown good performance and feasible implementation in both cases. © 2015 IEEE. The development of intelligent home security robot

No abstract