Fazrul Faiz Zakaria
Thumbnail Image
Preferred name
Fazrul Faiz Zakaria
Official Name
Fazrul Faiz, Zakaria
Alternative Name
Zakaria, Fazrul Faiz
Zakaria, F. F.
Main Affiliation
Scopus Author ID
55193708600
Researcher ID
GVS-6292-2022

Research Output

Filters

4 1
10
Reset filters

Settings
Now showing 1 - 10 of 10
Thumbnail Image
Publication

Traffic engineering provisioning of multipath link failure recovery in distributed SDN controller environment

2024-02-08 , Kelian V.H. , Mohd Warip M.N. , R Badlishah Ahmad , Ehkan P. , Fazrul Faiz Zakaria , Mohd Zaizu Ilyas

A revolutionary networking technology called Software-Defined Networking (SDN) enables better networking flexibility. In contrast to the conventional network, it provides another option for network development. SDN is characterized by the separation of the control and data planes in network architecture, implementation, and management. The central component of the network is the controller, which constitutes the control plane. The appropriate selection of a controller, along with determining the number and placement of controllers, plays a crucial role in optimizing resource utilization and guaranteeing network availability and network performance. Since SDN is still in its beginnings of development, it is virtually certain that further study will be needed in areas like design, particularly on the control plane, since the architecture directly affects the network's total performance. Furthermore, despite its intended purpose of managing networks on a large scale, SDN still presents challenges in effectively addressing network dynamics, such as the occurrence of link failures. This study presents a concept for the implementation of an SDN architecture. The proposed approach involves utilizing an Open Network Operating System (ONOS) open-source distributed SDN controller. The purpose of this implementation is to analyze network performance metrics and assess network availability. This study investigates the distributed SDN controller's performance on different scale networks: NSF, AEON, and TM topologies. Several metrics have been analyzed, including throughput, link failure detection, and Round-Trip-Time (RTT). The experiments use Mininet for emulation and Wireshark for real-time data packet capture and analysis. According to the study results, there is a positive correlation between network design complexity and controller load. The experiment emphasizes the resilience of distributed controllers, such as ONOS, in effectively recovering from link failures. This research will help academics and businesspeople who use distributed SDN controllers choose a controller and evaluate its effectiveness on the analyzed network architectures.

View
Thumbnail Image
Publication

Deep-Learning Assisting Cerebral Palsy Patient Handgrip Task Translation

2021-07-26 , Fazrul Faiz Zakaria , Mohd Nazri Mohd Warip , Phaklen Ehkan , Muslim Mustapa , Mohd Zaizu Ilyas

An electro-encephalography (EEG) brain-computer interface (BCI) can provide the brain and external environment with separate information sharing and control networks. EEG impulses, though, come from many electrodes, which produce different characteristics, and how the electrodes and features to enhance classification efficiency have been chosen has become an urgent concern. This paper explores the deep convolutional neural network architecture (CNN) hyper-parameters with separating temporal and spatial filters without any pre-processing or artificial extraction processes. It selects the raw EEG signal of electrode pairs over the cortical area as hybrid samples. Our proposed deep-learning model outperforms other neural network models previously applied to this dataset in training time (∼40%) and accuracy (∼6%). Besides, considerations such as optimum order for EEG channels do not limit our model, and it is patient-invariant. The impact of network architecture on decoder output and training time is further discussed.

View
Thumbnail Image
Publication

Deep Learning with FPGA: Age and Gender Recognition for Smart Advertisement Board

2023-10-06 , Yeoh W.S. , Fazrul Faiz Zakaria , Mustapa M. , Mohd Nazri Mohd Warip , Phak Len Al Eh Kan , Mozi A.M.

Age and gender recognition are helpful in various applications, especially in the field of advertising. To replace the traditional advertising method that can only display the same contents to all audiences, a smart advertisement board capable of detecting age and gender of audiences to display relevant contents is required to increase the effectiveness of advertising. This paper will use two image datasets to train and test the Convolutional Neural Network (CNN) based architecture models for age and gender recognition using deep learning. The dataset that produced the best performing model will be implemented on three different devices to observe the performance of the models on each device. A gender recognition model with accuracy of 91.53% and age recognition model with accuracy of 59.62% is produced. The results have also shown the use of Field Programmable Gated Array (FPGA) has greatly boosted the performance of the models in terms of throughput and latency.

View
No Thumbnail Available
Publication

A Proposal of Low Cost Home Automation System Using IoT and Voice Recognition

2020-03-20 , Phak Len Al Eh Kan , Keraf N.D. , Kelian V.H. , Fazrul Faiz Zakaria , Bei, Sin Zhen

Home Automation System is becoming more popular day by day due to its numerous benefits. This project proposes an idea in the design of low cost home automation system by using the Internet of Things (IoT) and voice recognition. The layout of the home divided into four areas and each area has own function and system. The Raspberry Pi 3 (RPi) Model B+ used as the main controller for the processing and transmitting the input data. IoT provided huge storage for data collection from sensors and home appliances. An Android application is developed to monitor the home environment and remotely control the home devices by using the button or voice. The speaker-independent recognition system by using Google Voice to Text on Android embedded in this project for physically challenged people to control the electrical appliances without moving. All the data will be stored in Firebase and can be retrieved at any time by the application and the RPi board. There is a side view of a prototype model with two floors and divided into four home areas. This Low-Cost Home Automation System using IoT and Voice Recognition is successfully achieved the project's objective.

View
Thumbnail Image
Publication

Perceptual-based features for blind image quality assessment using extreme learning machine for biodiversity monitoring

2024-02-08 , Verak N. , Pakhlen Ehkan , Ruzelita Ngadiran , Jungjit S. , Fazrul Faiz Zakaria , Mohd Nazri Mohd Warip , Ilyas M.Z.

Blind Image Quality Assessment (BIQA) is crucial for various image processing applications, including image denoising, transmission evaluation, optimization, watermarking, and situations where a reference image is unavailable. However, existing state-of-the-art Image Quality Assessment (IQA) metrics are often specific to certain types of distortion and fail to align with human perception. To address this issue, our research study proposes a novel approach that incorporates perceptual-based features and utilizes a pooling algorithm based on Extreme Learning Machine (ELM). By considering human visual characteristics and the impact on image content, we aim to mimic human perception, which can detect noticeable differences at specific frequency ranges, akin to neurons. The work is divided into three phases. In the first phase, we derive perceptual features using lifting wavelet, focusing on texture, edge, and contrast components. Subsequently, in the second phase, these features are trained to generate output for pooling using Extreme Learning Machine (ELM). The pooling strategy of ELM is chosen due to its ability to overcome limitations found in previous pooling techniques like Neural Networks (NNs) and Support Vector Regression (SVR). This approach enables us to evaluate the quality score of images accurately, benefiting from ELM's superior learning accuracy and faster learning speed. The third phase involves performance evaluation, including statistical analysis for algorithm validation and a comparison with existing BIQA methods using MATLAB software. We verify our proposed approach on diverse image databases containing various distortion types, aiming to create a general-purpose BIQA solution. The outcomes of this work will have significant implications in several image processing applications, such as optimizing image enhancement for medical purposes like tumor or cancer detection, image watermarking for security applications, image coding and compression, and image forensic analysis. In biodiversity monitoring, image enhancement plays a crucial role in tracking and data collection.

View
Thumbnail Image
Publication

Breast cancer classification using deep learning and FPGA inferencing

2023-02-21 , Wong E.H. , Fazrul Faiz Zakaria , Mustapa M. , Mohd Nazri Mohd Warip , Phak Len Al Eh Kan

Implementing deep learning technology with FPGA as an accelerator has become a popular application due to its efficiency and performance. However, given the tremendous data generated on medical diagnosis, normal inference speed is not sufficient. Hence, the FPGA technology is implemented for fast inference. In this context, the FPGA accelerates the deep learning inference process for fast breast cancer classification with minimal latency on real-time deployment. This paper summarizes the findings of model deployment across various computing devices in deep learning technology with FPGA. The study includes model performance evaluation, throughput, and latency comparison with different batch sizes to the extent of expected delay for real-world deployment. The result concludes that FPGA is the most suitable to act as a deep learning inference accelerator with a high throughput-to-latency ratio and fast parallel inference.

View
Thumbnail Image
Publication

Compact Meandered Monopole Antenna for Dual-Bands WLAN Application

2021-07-26 , Bohari S. , Faudzi N.M. , Razali A.R. , Mozi A.M. , Fazrul Faiz Zakaria , Mohd Zaizu Ilyas

A compact meandered dual-bands monopole antenna for the application of a wireless local area network (WLAN) is proposed. This antenna has two operating frequency bands which are 2.4 GHz and 5.2 GHz and denoted as lower and upper operating bands respectively. In the antenna design, a meandered arms structure has been proposed to obtain a compact size monopole antenna with an overall dimension of 30 mm x 21 mm. Furthermore, the dual-bands operating frequency is achieved with the contribution of two meandered arms structure as well as a partial ground plane proposed in the antenna design. The copper layer traces with the thickness of 0.035 mm has been used as the radiating patch and the partial ground plane has been printed at the back side of the FR-4 substrate with the permittivity, ϵ_r of 4.5 and the thickness of 1.6 mm. The proposed antenna has a simple design, small size, easy to fabricate and low cost. The measured and simulated results were compared to analyse the performance of the designed antenna. From the simulation, the operating frequencies achieved are at 2.44 GHz and 5.23 GHz, while from the measurement at 2.50 GHz and 4.44 GHz. Other antenna parameter such as radiation pattern and gain has also been evaluated and analysed.

View
Thumbnail Image
Publication

Comparative Study of Parallelism and Pipelining of RGB to HSL Colour Space Conversion Architecture on FPGA

2020-03-20 , Pakhlen Ehkan , Siew, Soon Voon , Fazrul Faiz Zakaria , Mohd Nazri Mohd Warip , Mohd Zaizu Ilyas

RGB colour model is a basic colour model and complements together to produce full colour range but it is unable to produce sufficient information for digital image analysis. However, HSL is capable to provide other useful information such as colour in degree, saturation of the colour and brightness of colour. In this work, RGB to HSL mathematical conversion algorithm is implemented on FPGA chip. Parallelism and pipelining capabilities of FPGA helps to speed up conversion performance. The RGB to HSL equation is implemented by using two architectures which are parallel and 7-stages pipeline architectures. The designed parallel and pipeline converters have one clock and seven clock cycle of data latency respectively. The parallel and pipeline architectures for RGB to HSL converter have been achieved rate of accuracy by hardware verification up to 99% and 98% and possessed maximum operating frequency merit of 50 MHz and 120 MHz respectively.

View
Thumbnail Image
Publication

Tuberculosis Classification Using Deep Learning and FPGA Inferencing

2023-02-01 , Fazrul Faiz Zakaria , Asral Bahari Jambek , Norfadila Mahrom , Rafikha Aliana A Raof , Mohd Nazri Mohd Warip , Al Eh Kan P.L. , Mustapa M.

Among the top 10 leading causes of mortality, tuberculosis (TB) is a chronic lung illness caused by a bacterial infection. Due to its efficiency and performance, using deep learning technology with FPGA as an accelerator has become a standard application in this work. However, considering the vast amount of data collected for medical diagnosis, the average inference speed is inadequate. In this scenario, the FPGA speeds the deep learning inference process enabling the real-time deployment of TB classification with low latency. This paper summarizes the findings of model deployment across various computing devices in inferencing deep learning technology with FPGA. The study includes model performance evaluation, throughput, and latency comparison with different batch sizes to the extent of expected delay for real-world deployment. The result concludes that FPGA is the most suitable to act as a deep learning inference accelerator with a high throughput-to-latency ratio and fast parallel inference. The FPGA inferencing demonstrated an increment of 21.8% in throughput while maintaining a 31% lower latency than GPU inferencing and 6x more energy efficiency. The proposed inferencing also delivered over 90% accuracy and selectivity to detect and localize the TB.

View
Thumbnail Image
Publication

Toward Adaptive and Scalable Topology in Distributed SDN Controller

2023-03-01 , Kelian V.H. , Mohd Nazri Mohd Warip , R Badlishah Ahmad , Phak Len Al Eh Kan , Fazrul Faiz Zakaria , Mohd Zaizu Ilyas

The increasing need for automated networking platforms like the Internet of Things, as well as network services like cloud computing, big data applications, wireless networks, mobile Internet, and virtualization, has driven existing networks to their limitations. Software-defined network (SDN) is a new modern programmable network architectural technology that allows network administrators to control the entire network consistently and logically centralized in software-based controllers and network devices become just simple packet forwarding devices. The controller that is the network's brain, is mostly based on the OpenFlow protocol and has distinct characteristics that vary depending on the programming language. Its function is to control network traffic and increase network resource efficiency. Therefore, selecting the right controllers and monitoring their performance to increase resource usage and enhance network performance metrics is required. For network performance metrics analysis, the study proposes an implementation of SDN architecture utilizing an open-source OpenDaylight (ODL) distributed SDN controller. The proposed work evaluates the deployment of distributed SDN controller performance on three distinct customized network topologies based on SDN architecture for node-to-node performance metrics such as delay, throughput, packet loss, and bandwidth use. The experiments are conducted using the Mininet emulation tool. Wireshark is used to collect and analyse packets in real-time. The results obtained from the comparison of networks are presented to provide useful guidelines for SDN research and deployment initiatives.

View