Searching for parking in large cities can be a painful experience. The existing technologies that provide parking information based on sensors are very costly and limited. Drivers do not have access to real-time accurate parking information such as nearest parking locations and available parking spots. The objective of this research is to develop a low-cost vehicle counting device using IoT technology to record vehicles’ entry and exit time, the real-time parking availability information can be derived from the recorded data and disseminated to the drivers. Effective methods for traffic congestion prediction will also be formulated based on parking space availability. This research will be divided into two parts: (1) Development of low-cost IoT device and cloud architecture, (2) Formulation of traffic congestion a method using deep Q-learning framework. IoT devices will be designed to collect real-time vehicle entry and exit data and sent it to the cloud. The collected data will consist of time-in and out of the vehicles for each day. Data analysis will be performed to identify patterns in the parking data. The patterns will be used to make real-time predictions regarding the availability of parking in the area. In some cases, the parking data could be random. Trending and de-trending techniques will be employed along with the time series analysis to separate the deterministic part of the data from random parts. A Deep Q-learning strategy will be explored and compared to other deep learning methods to provide a better traffic congestion forecast in selected areas. A Master’s student will be trained for this research. One Scopus and one WoS indexed journal will also be produced. The expected outcome is a real-time parking system to provide parking availability prediction and improve the traffic in the city that could realize the smart parking criteria in the MyDIGITAL blueprint.

michael.goh@mmu.edu.my

Both IQA and IAA share some common underlying factors that affect user judgments, but existing works ignore this correlation in predicting image quality. In this research, the objective is to study the correlation between IQA and IAA, and design a novel machine learning model to jointly predict the technical and aesthetics quality of an image. the AI model of this research has various potential applications including image retrieval, image restoration, in-camera guide, robot photography, and image cropping or thumbnailing.

Note: The scope of the project can be widened to that of a Ph.D. degree.

lkwong@mmu.edu.my

An image can invoke various emotions, depending on the visual features and semantic content of the image. Notably, similar image content in different contexts might induce different emotions. Eg. If we see a famous football player crying on his knees, the audience may feel sad; but if this is after winning a game, the audience may feel excited. However, most of the existing image emotion prediction models ignore the context of the image or did not focus much on extracting the image semantics that can help to improve the prediction of image emotion. In this research, the key objective is to build a model that perform context-aware emotion prediction from images using deep learning techniques. The developed algorithm can be very useful for many applications, including image retrieval, intelligent billboards, media content curation, web design, scene-aware music synthesis, and sentiment analysis.

Note: The scope of the project can be widened to that of a Ph.D. degree.

lkwong@mmu.edu.my

Gait analysis, as a non-intrusive method for biometric identification, has garnered attention across various fields including security systems, healthcare monitoring, and forensic science. However, its potential in predicting demographic attributes such as age and gender remains relatively untapped. This study delves into the feasibility of predicting age and gender through human gait features. Relevant features will be extracted from the gait patterns and machine learning algorithms will be employed to develop predictive models.

tee.connie@mmu.edu.my

This research delves into the utilization of gait behavior as a screening tool for Parkinson’s disease (PD). Parkinson’s disease is a neurodegenerative disorder characterized by a spectrum of motor symptoms, including gait abnormalities, which often manifest even in the early stages of the disease. Leveraging advancements in technology, particularly in motion analysis and machine learning, this study investigates the potential of gait behavior analysis as a non-invasive and cost-effective method for PD screening.

tee.connie@mmu.edu.my

In Malaysia, medical records in the clinics are entered by the doctors according to their style and preference. Some write the prescription using short-form other may add additional notes and diagnostics information. If these data can be cleaned, extracted and store in a systematic and structured manner, then these data can be used to generate meaningful insight for pharmacy, hospital and doctors to know the effectiveness of the drugs given to the patients and understand the demand of the drugs according to location and area. The idea of this project is to conduct study and perform analysis on a large prescription data provided by clinics, develop algorithms using deep learning approach to extract and structured the prescription data so that it can be used to generate meaningful insight. The developed algorithm is required to compare against the existing solutions.

michael.goh@mmu.edu.my

Translation, rotation and scale invariants (TRSI) of Krawtchouk moments proposed are well-known local feature extractors. However, existing invariants are computationally intensive and failed to resolve spatial deformations on weight function in the invariant function. As such, invariants generated by present algorithms are non-orthogonal, which compromise its discriminative power. This project aims to simplify and improve the computational efficiency of the TRSI algorithm for Krawtchouk moments. New normalization schemes with adaptive weight function will be formulated to produce high discriminative local features.

Agriculture sector often suffers great losses due to plant diseases. Image-based classification of plant diseases holds potential for early plant disease detection. However, automatic plant condition detection faces many challenges associated with large variations of visual symptoms, background and illumination. In this work, we proposed a moment-based Deep Neural Network (DNN) that utilizes our improved, high discriminative Krawtchouk moments for improving plant condition classification.

This project begins by formulating recursive functions of affine Krawtchouk transformation for fast computation, followed by the formulation of TRSI normalization schemes and modification of the weight function for feature extraction. Next, plant images of healthy and unhealthy plants will be collected, and preprocessed. Region-of-interests (ROI) for each image will then be identified and TRSI is calculated to extract features from ROIs. Finally a DNN model is designed and trained to classify plant health conditions.

cypee@mmu.edu.my

Weather and other factors (like COVID-19) often prevent people from going to exercise outside. Standard modern smartphones have step trackers that will track walking, even indoors, but most exercise gamification involves being able to go outdoors. We want to explore what game features would encourage people to keep exercising even when indoors.

Note: This is a wide open area and could be expanded to a Ph.D.

ianchai@mmu.edu.my

Problem Statement: Time series and streaming data are generated every minute and second due to the available sensors and software in Smart Home Systems. With a combination of The Internet of Things (IoT) and Smart Home Energy Management Systems (SHEMS), it becomes a scorching topic nowadays for a better process of extracting useful knowledge. It helps for better management and visualization of electricity usage. However, there are few challenges faced by the system developed, such as data quality or data anomaly issues. These anomalies can be due to technical or non-technical faults. The non-technical fault is essential to detect, which might cause economic cost. It also becomes challenging to train models in the case of an unlabeled dataset.

Objective: The main objective is to find data abnormalities in time-series data for better monitoring and forecasting. Secondly, the system should able to distinguish between actual anomaly and seasonal anomaly.

Methodology: Here an intelligent system will be developed using unsupervised learning and integrate into the smart home system. The model will be capable of clustering data into anomalies and not. It will automatically adapt to changing values in different environments. For determining seasonal anomalies, feature engineering will be processing before model training.

Expected Outcome: Finally, it is expected that the developed anomaly detector will be able to detect all anomalies as soon as possible, triggering real alarms in real-time for time-series data’s energy consumption or electricity theft. The users will be able to interact with the intelligent model from their browser. Even the developers will be able to ingrate the mode in their system using API.

Conclusion: Detecting data anomalies will help make a better decision to reduce energy usage wasted or point out if someone is stealing electricity.

jakir.hossen@mmu.edu.my

This project is about dynamic handwritten signature recognition. It is expected to propose a feature extraction approach to analyze the dynamic handwritten signature based on spatial and temporal attributes.

yhpang@mmu.edu.my

This project is about processing EEG signal for Alzheimer’s patients classification. The student is expected to explore

1. Preprocessing of EEG signals,
2. Feature extraction of the EEG signals, and
3. Classification on the extracted features.

Supervised based feature extraction approach is a good alternative to well describe the signal patterns.

yhpang@mmu.edu.my

Human activity recognition is the problem of classifying sequences of inertial data recorded by specialized harnesses or smart phones into known well-defined movements.

yhpang@mmu.edu.my

Undeniable signature scheme is a special featured signature scheme whereby the validity and invalidity of an undeniable signature can only be verified with the signer’s consent by engaging in a confirmation protocol and a disavowal protocol respectively, as opposed to a digital signature which is universally verifiable. Some potential applications of undeniable signatures are such as electronic payment, electronic voting, electronic auctions, etc. This research focuses on the analysis of relationships among the security properties of undeniable signature scheme and its potential applications.

shheng@mmu.edu.my

This study focuses on the review and analysis of the existing blockchain-based cryptographic applications, and proposes new applications. The new cryptographic applications will be supported with implementation and performance and comparison analysis, and rigorous theoretical framework if applicable.

shheng@mmu.edu.my

Car parking with good security system is more preferable with those cars parking who does not installed with any security system. One of the benefits of car parking system with good security is that it able to reduce the risk of car theft. The better the security system, the lower the risks the car theft would able to steal vehicles. Car parking with good security system are usually installed with CCTV and often hire security guards to patrol the car parking area. Although the risk of car burglar cannot be completely eliminated, but at least it is able to reduce more risk of a vehicle being stolen compare to the park at your own risk car parking system. In this research, we propose to build up a visual car burglar detection system using wireless webcam with Car burglar detection algorithm with automatically identify/ monitor if there is any possible car burglar case occur. This research will develop good helper in surveillance and evidence keep for car park security.

Note: For candidates with electronics background and has interest in image processing.

wkwong@mmu.edu.my

This research required to come out an effective automated fever detection method in thermal imaging system. It utilizes deep learning approach for human object detection and automated fever detection algorithm to be applied in thermal imaging system on contactless human body temperature measurement.

Note: For candidates with electronics background and has interest in image processing.

wkwong@mmu.edu.my

Agriculture sector often suffers great losses due to plant diseases. Image-based classification of plant diseases holds potential for early plant disease detection. However, automatic plant condition detection faces many challenges associated with large variations of visual symptoms, background and illumination. In this work, we proposed a moment-based Deep Neural Network (DNN) that utilizes our improved, high discriminative Krawtchouk moments for improving plant condition classification.

cypee@mmu.edu.my

Recent studies have shown that the trend of suicide is approaching younger aged citizens, which is reported as below the age of 40. Many concerns have been raised by parents and guardians about the impact of social media such as Facebook, Instagram, etc. on teenagers, in which tons of misleading information can be easily accessed by them at any time of the day. The advance in communication technologies nowadays could cause a greater mental health crisis among youngsters in the near future. This research project proposes to develop a real-time prototype suicide prediction system. The predictive model is trained based on voice recognition and analysis using machine learning. The proposed system can be used for preventing possible suicide activities, which includes the functions of mental health monitoring, early sign of suicide detection as well as emergency notification through the Internet of Things (IoT) platform.

gcchung@mmu.edu.my

Driver behaviour detection and evaluation is becoming an essential task for vehicle manufacturers. Driver distraction is the major cause of road accidents and infrastructure deformation. Furthermore, secondary roads accidents are mainly affected, since external distraction and pedestrian presence are higher than highways. This research proposes a comparison of several machine learning classification methods to identify the driver’s behaviour on secondary roads. The classification and comparison are based on the evaluation of real data.

gcchung@mmu.edu.my

The project aims to develop computer-aided diagnosis system for nasopharyngeal carcinoma using machine learning and image processing.

faizal1@mmu.edu.my

The DiSAR is a new approach of conducting roadshow. The model leverages the digital signages to improve the roadshow experience. The research explores multiple directions on how to utilize the digital signages to achieve a better roadshow efficiency and effectiveness. The possible research areas are the possible interaction modalities and their usability; the overall flow of the DiSAR roadshow and how to optimize the traffic flow; the data analytics on various dimensions of the DiSAR; machine learning and AI techniques on optimizing the DiSAR (ie. prediction on user behavior, participant profiling etc.).

mcleow@mmu.edu.my

This research focuses on addressing security and privacy concerns in Vehicular Communication Networks (VCNs) within the Internet of Vehicles (IoV). The thesis will explore the application of blockchain technology to establish a secure and trustworthy communication framework for connected vehicles. The study aims to design and implement a decentralized and tamper-resistant system that ensures data integrity, privacy protection, and secure communication among vehicles and with infrastructure. The research will contribute to enhancing the overall security posture of IoV ecosystems.

msaberik@mmu.edu.my

This research investigates the use of blockchain technology to establish a decentralized and secure health data exchange platform in the IoMT. The thesis aims to design and implement a system that ensures the integrity, traceability, and security of medical data shared among connected devices. The study contributes to building a trustworthy infrastructure for medical information exchange within IoMT ecosystems.

msaberik@mmu.edu.my

This thesis focuses on the implementation of machine learning algorithms for early disease detection in the context of IoMT-based remote patient monitoring. The research aims to develop predictive models that analyze data from wearable medical devices to identify patterns indicative of health issues. The study contributes to advancing the capabilities of IoMT in proactive healthcare management.

msaberik@mmu.edu.my