Kah Phooi Seng

Radial Basis Function Neural Network With Incremental Learning for Face Recognition

Conventional face recognition suffers from problems such as extending the classifier for newly added people and learning updated information about the existing people. The way to address these problems is to retrain the system which will require expensive computational complexity. In this paper, a radial basis function (RBF) neural network with a new incremental learning method based on the regularized orthogonal least square (ROLS) algorithm is proposed for face recognition. It is designed to accommodate new information without retraining the initial network. In our proposed method, the selection of the regressors for the new data is done locally, hence avoiding the expensive reselecting process. In addition, it accumulates previous experience and learns updated new knowledge of the existing groups to increase the robustness of the system. The experimental results show that the proposed method gives higher average recognition accuracy compared to the conventional ROLS-algorithm-based RBF neural network with much lower computational complexity. Furthermore, the proposed method achieves higher recognition accuracy as compared to other incremental learning algorithms such as incremental principal component analysis and incremental linear discriminant analysis in face recognition.Conventional face recognition suffers from problems such as extending the classifier for newly added people and learning updated information about the existing people. The way to address these problems is to retrain the system which will require expensive computational complexity. In this paper, a radial basis function (RBF) neural network with a new incremental learning method based on the regularized orthogonal least square (ROLS) algorithm is proposed for face recognition. It is designed to accommodate new information without retraining the initial network. In our proposed method, the selection of the regressors for the new data is done locally, hence avoiding the expensive reselecting process. In addition, it accumulates previous experience and learns updated new knowledge of the existing groups to increase the robustness of the system. The experimental results show that the proposed method gives higher average recognition accuracy compared to the conventional ROLS-algorithm-based RBF neural network with much lower computational complexity. Furthermore, the proposed method achieves higher recognition accuracy as compared to other incremental learning algorithms such as incremental principal component analysis and incremental linear discriminant analysis in face recognition.

Low memory image stitching and compression for WMSN using strip-based processing

Due to the limited Field-Of-View (FOV) of a single camera, it is sometimes desired to extend the FOV using multiple cameras. Image stitching is one of the methods that can be used to exploit and remove the redundancy created by the overlapping FOV. However, the memory requirement and the amount of computation for conventional implementation of image stitching are very high. In this paper, this problem is resolved by performing the image stitching and compression in a strip-by-strip manner. First, the stitching parameters are determined by transmitting two reference images to an intermediate node to perform the processing. Then, these parameters are transmitted back to the visual node and stored in there. These parameters will be used to determine the way of stitching the incoming images in a strip-by-strip manner. After the stitching of a strip is done, it can be further compressed using a strip-based compression technique.

Lane Detection and Kalman-Based Linear-Parabolic Lane Tracking

This paper presents a lane detection and linear-parabolic lane tracking system using kalman filtering method. First, the image horizon is detected in a traffic scene to split the sky and road region. Road region is further analyzed with entropy method to remove the road pixels. Lane boundaries are then extracted from the region using lane markings detection. These detected boundaries are tracked in consecutive video frames with a linear-parabolic tracking model. The model parameters are updated with Kalman filtering method. Error-checking is performed iteratively to ensure the performance of the lane estimation model. Simulation results demonstrate good performance of the proposed Kalman-based linear-parabolic lane tracking system with fine parameters update.

Intra color-shape classification for traffic sign recognition

This paper presents a novel traffic sign recognition system comprising of: (i) Color/shape classification, (ii) Pictogram extraction, (iii) Features selection and, (iv) Lyapunov Theory-based Radial Basis Function neural network (RBFNN). In the proposed system, traffic signs are first segmented and classified with regard to its unique color and shape in order to partition a large set of data into smaller subclasses. Within these subclasses, all redundant information except the pictogram is discarded for feature selection since the pictogram contains critical information for road users. Principle Component Analysis (PCA) is applied to extract salient points for traffic sign dimensionality reduction. This is followed by the Fishers Linear Discriminant (FLD) to further obtain the most discriminant features. These features are fed into RBFNN for training with a proposed weight updating scheme based on Lyapunov stability theory. The performance of the proposed system is evaluated with Malaysian road signs with promising recognition rate.

Bottom-up visual saliency map using wavelet transform domain

Object or region based image processing can be performed more efficiently with information pertaining locations that are visually salient to human perception with the aid of a saliency map. The saliency map is a master topological map having the possible locations of objects or regions which a human perceived as important/salient. In this paper, a method to compute the saliency map in the wavelet transform domain is explored. Previous works involving saliency in this domain usually involves salient points, which are in fact accurate but they do not cover the area as a region and involve heavy repeated iterations. The method explored in this paper is compared to two state-of-art methods in which these methods involve the frequency domain. The presented method provides more accurate salient regions compared to the other two methods while retaining a resolution which the salient regions are visually identifiable.

Dual optimal multiband features for face recognition

Illumination and expression variations degrade the performance of a face recognition system. In this paper, a novel dual optimal multiband features method for face recognition is presented. This method aims to increase the robustness of face recognition system to both illumination and expression variations. The wavelet packet transform decomposes image into frequency subbands and the multiband feature fusion technique is incorporated to select optimal multiband feature sets that are invariant to illumination and expression variation separately. Parallel radial basis function neural networks are employed to classify the two sets of feature. The scores generated are then combined and processed by an adaptive fusion mechanism. In this mechanism, the level of illumination variations of the input image is estimated and the weights are assigned to the scores accordingly. Experiments based on Yale, YaleB, AR and ORL databases show that the proposed method outperformed other algorithms.

Real-Time Implementation of Vision-Based Lane Detection and Tracking

This paper presents a real-time implementation on lane detection and tracking system in order to localize lane boundaries and estimate a linear-parabolic lane model. It is realized using TMS320DM642 DSP board. Video frame is first captured with CCD camera and stored in video port buffer. Next, input image is split into sky and road region with horizon localization. Lane analysis is applied on the road region to remove road pixels. Only lane markings are the interests for the lane detection process. Once lane boundaries are located, the possible edge pixels are scanned to continuously obtain the lane model. Linear-parabolic model is used to construct the geometry of the lane. The model parameters are updated with Kalman filtering. Video sequences are tested to verify the performance of the system and it has good performance.

Wireless intelligent incontinence management system using smart diapers

This paper presents a wireless intelligent incontinence management system being developed for the University Malaya Medical Center (UMMC) that utilizes ldquosmartrdquo diapers to discreetly monitor and estimate wetness events, detect other relevant clinical conditions and alert staff by transmitting information via wireless technology to an intelligent central management system. We describe the hardware and software modules of the system and give results conducted on diaper experiments and the transmission range of the wireless technology.

Lips Contour Detection and Tracking Using Watershed Region-Based Active Contour Model and Modified

In this paper, a region-based active contour model (ACM) with local information using watershed segmentation is proposed for lips contour detection. Compared to the ACM with global energy terms, the proposed system provides a more precise lips contour convergence under the circumstances where the lips are difficult to distinguish using global statistics. Furthermore, since the ACM is sensitive to the initial contour position, a modified H∞ based on Lyapunov stability theory is proposed to provide better tracking of the subsequent lips feature points as the ACM initialization. The integration of the proposed ACM and modified H∞ has revealed an improvement of the overall lips contour detection.

H_{\infty}

In this paper, a new multimodal biometric recognition system based on feature fusion is proposed to increase the robustness and circumvention of conventional multimodal recognition system. The feature sets originating from the output of the visual and audio feature extraction systems are fused and being classified by RBF neural network. Other than that, 2DPCA is proposed to work in conjunction with LDA to further increase the recognition performance of the visual recognition system. The experimental result shows that the proposed system achieves a higher recognition rate as compared to the conventional multimodal recognition system. Besides, we also show that the 2DPCA+LDA achieves a higher recognition rate as compared with PCA, PCA+LDA and 2DPCA.