Noridayu Manshor

Multi-feature Based Object Class Recognition

In object class recognition, lots of past researches focused on the local descriptors such as SIFT to categorize the variation of objects belonging to the same category in different poses, sizes, and appearance. However, SIFT descriptors may produce poor result especially if the object does not have enough information of its texture features. Due to this problem, we hypothesize that the use multi feature may increase the performance of object class recognition. In this paper, we use additional global shape features, Fourier Descriptors combined with SIFT descriptors to help in improving the performance of object class recognition. The selection of shape features is chosen due to the objects are easier to describe based on this features from human perspective compare to other features. We have divided our experiments into two: Experiment E1 is limited to the side view of bike, car, horse, and cow images whereas Experiment E2 consists of similar categories of dataset but in arbitrary views, rotations, and scales. The dataset we used in our experimentation are obtained from PASCAL, Weizmann and TU Darmstadt database. We assume that all objects are segmented manually before the feature extraction process. We validate our selection features using K-Means algorithm to evaluate the features for the purpose of object class recognition. Our results indicate that the combination of additional shape features together with SIFT descriptors performs better than using SIFT descriptors alone by up to 15% with limitation views of images.

Performance evaluation of time-based black hole attack detection in mobile ad hoc networks

Efficient and quick attack detection is critical in any networks, especially if the attack is harmful and can bring down the whole network within a short period of time. A black hole or packet drop attack is one example of a harmful attack in mobile ad hoc networks. In this study, we implement a series of time-based black hole attack detection of different time intervals and compare the results. We study the performances of the networks, the packet delivery ratio percentage, with detection interval time of 900, 450 and 300 seconds with a total of 900 seconds of simulation time. The results suggest that appropriate time interval is critical in providing reliable detection results in timely manner. In general, the 450 seconds detection interval time has provided more reliable results, with lower false positive percentage in comparison to those of the 300 seconds detection interval time. The best explanation to the high false positive rate in the shorter detection interval time is due to the insufficient time given to the packets to arrive to the destinations during the detection process. Meanwhile, implementing the attack detection only after 900 seconds may be considered too late and thus, may have a devastating impact to the networks.

Object Based Clustering Using Hybrid Algorithms

This paper presents a hybrid algorithm for object based clustering. The algorithm is designed based on hybrid of hierarchical and k-means clustering algorithm. For this work, we used dataset consist of natural imagery collected from PASCAL database 2006 collection and Google images. A collection of low level features image is used to validate the performance of our approach. Experimental results show that hybrid algorithm produced higher accuracy compared to k-means and hierarchical algorithm by up to 25%.

Dyslexia Adaptive Learning Model: Student Engagement Prediction Using Machine Learning Approach

Education barriers are synonym with people with dyslexia life experience. People with dyslexia encounter barriers such as in academic related areas, mistreated with negative reaction on their behaviour and limitation to acquire a suitable support to overcome the barriers. Therefore, this work focus on giving the support to help students with dyslexia deal with their difficulty through adaptively sense their behaviour for engagement perspective. For that reason, we apply machine learning approach that utilises Bag of Features (BOF) image classification to predict student engagement towards the learning content. The engagement prediction was relatively using frontal face of the 30 students. We used Speeded-Up Robust Feature (SURF) key point descriptor and clustered using k-Means method for the codebook in this BOF model. Then, we classify the model using 3 types of classifier which are Support Vector Machine (SVM), Naive Bayes and K-Nearest Neighbour (k-NN) to find the best classification result. Through these methods, we managed to get high accuracy with 97–97.8%.

Engagement Prediction in the Adaptive Learning Model for Students with Dyslexia

Student engagement is one of the most important elements in a likelihood of school failure or dropout. Therefore, it is vital to measure the student engagement as quickly as possible and as often as possible to prevent it occurred in a prolonged situation. There a few ways to assess the engagement that includes self-reporting, teachers rating, interviews and observation. However, these methods are not only takings time but also need a lot of hard work, cost and difficult to conduct for a very short time. Therefore, we a proposing an alternative to predict student engagement through frontal face detection. We apply machine learning approach that utilizes Speed-Up Robust Features (SURF) descriptor to detect key interest point of the images and cluster using different codebook sizes. For classification model, we used Support Vector Machine (SVM) with two different kernels and Naïve Bayes. We managed to get more than 88% of the accuracy results. The model is an important part of our proposed adaptive learning model for dyslexic students.

Food Category Recognition Using SURF and MSER Local Feature Representation

Food object recognition has gained popularity in recent years. This can perhaps be attributed to its potential applications in fields such as nutrition and fitness. Recognizing food images however is a challenging task since various foods come in many shapes and sizes. Besides having unexpected deformities and texture, food images are also captured in differing lighting conditions and camera viewpoints. From a computer vision perspective, using global image features to train a supervised classifier might be unsuitable due to the complex nature of the food images. Local features on the other hand seem the better alternative since they are able to capture minute intricacies such as interest points and other intricate information. In this paper, two local features namely SURF (Speeded- Up Robust Feature) and MSER (Maximally Stable Extremal Regions) are investigated for food object recognition. Both features are computationally inexpensive and have shown to be effective local descriptors for complex images. Specifically, each feature is firstly evaluated separately. This is followed by feature fusion to observe whether a combined representation could better represent food images. Experimental evaluations using a Support Vector Machine classifier shows that feature fusion generates better recognition accuracy at 86.6%.

Informing Design of an Adaptive Learning Model for student with Dyslexia: A Preliminary Study

Students with dyslexia are known to have difficulties in phonology, spelling, reading, and writing. Therefore, specific intervention needs to be introduced to the students in order to help overcome their difficulties. The existence of Dyslexia Association of Malaysia (DAM) that provides dyslexic intensive education program becomes a primary place for parents to seek for help in intervention. Based on DAM experience in handling students with dyslexia, we conducted a preliminary study comprises semi-structured interview and observation result to uncover their teaching approaches and materials. The result from this preliminary study will be used to develop the adaptive learning model in order to make an effective learning experience that tailored to individual difficulties

An integration of minimum local feature representation methods to recognize large variation of foods

Local invariant features have shown to be successful in describing object appearances for image classification tasks. Such features are robust towards occlusion and clutter and are also invariant against scale and orientation changes. This makes them suitable for classification tasks with little inter-class similarity and large intra-class difference. In this paper, we propose an integrated representation of the Speeded-Up Robust Feature (SURF) and Scale Invariant Feature Transform (SIFT) descriptors, using late fusion strategy. The proposed representation is used for food recognition from a dataset of food images with complex appearance variations. The Bag of Features (BOF) approach is employed to enhance the discriminative ability of the local features. Firstly, the individual local features are extracted to construct two kinds of visual vocabularies, representing SURF and SIFT. The visual vocabularies are then concatenated and fed into a Linear Support Vector Machine (SVM) to classify the respective fo...

Impact of node inter-domain movement on MANETs performance

Packet delivery ratio (PDR) percentage is one of the important network performance indicators in MANETs. In general, the PDR value degrades as speed of the node increases and coupled with high mobility or constant movement. As more nodes move at high speed, more broken path or link breakage occur and thus, more packets will be dropped. Interestingly, PDR rate has also been used to detect packet drop or black hole attack in the network. Thus, the packet drop activity may due to either the broken path process itself or deliberate drop by malicious nodes. Validating the packet drop action itself is imperative in reducing the false positive rate during the attack detection. This paper studies the movements of nodes in the networks that have caused high packet drop percentage. In particular, we investigate the inter-domain movement since it has substantial effect on the packet drop percentage. To the best of our knowledge, this is the first work that studies such relationship. The results on the overall network show that the high number of inter-domain movement may not necessarily contribute significantly to the packet drop percentage. However, when focus is on the inter-domain movement of the critical nodes, we yield consistent results. The proposed monitoring approach is also energy efficient as it reduces the need to monitor other large number of nodes insignificant movements.

Fusion of Global Shape and Local Features Using Boosting for Object Class Recognition

In object class recognition, the state-of-the-art works shows using combination varies local features may produce a good performance in recognition. These local features may have a different performance on one category to other category which it depends on the richness of local features. Due to that limitation, the shape features of objects are taken into consideration to be combined with local features. In this paper, we use Fourier Descriptor (FD), Elliptical Fourier Descriptors (EFD) and Moment Invariant (MI) as a global shape feature and Scale Invariant Feature Transform (SIFT) as local features. For learning technique, boosting is used in improving the recognition objects. This approach identifies the correct and misclassified dataset iteratively. Experimental results indicate that the recognition model outperform improved the accuracy of classification by up to 10% that is comparable to or better than that of state-of-the-art approaches.