Department Publications

Machine-learning-assisted cancer subtyping is a promising avenue in digital pathology. Cancer subtyping models however require careful training using expert annotations, so that they can be inferred with a degree of known certainty (or uncertainty). To this end, we introduce the concept of uncertainty awareness into a self-supervised contrastive learning model. This is achieved by computing an evidence vector at every epoch, which assesses the model's confidence in its predictions. The derived uncertainty score is then utilized as a metric to selectively label the most crucial images that require further annotation, thus iteratively refining the training process. With just 1-10% of strategically selected annotations, we attain state-of-the-art performance in cancer subtyping on benchmark datasets. Our method not only strategically guides the annotation process to minimize the need for extensive labeled datasets, but …

Water resource management in Sri Lanka is hindered by inefficient water distribution systems in reservoirs and tanks, resulting in water scarcity, wastage, and negative impacts on agriculture, hydropower generation, and drinking water. Despite the global progress in Machine Learning (ML)-based predictive models for hydrological forecasting, Sri Lanka lacks the application of such techniques in its reservoir water management systems. This work proposes a novel approach that integrates ML and Internet of Things (IoT) technologies to develop a demand prediction and smart water distribution system tailored to the Sri Lankan context. The proposed system employs a hybrid model combining Artificial Neural Networks (ANN) and Long Short-Term Memory (LSTM) networks to accurately predict reservoir inflow, release, and water levels. Sensory data from ESP32 microcontrollers is collected and transmitted to a cloud …

Neonates are highly susceptible to seizures, often leading to short or long-term neurological impairments. However, clinical manifestations of neonatal seizures are subtle and often lead to misdiagnoses. This increases the risk of prolonged, untreated seizure activity and subsequent brain injury. Continuous video electroencephalogram (cEEG) monitoring is the gold standard for seizure detection. However, this is an expensive evaluation that requires expertise and time. In this study, we propose a convolutional neural network-based model for early prediction of neonatal seizures by distinguishing between interictal and preictal states of the EEG. Our model is patient-independent, enabling generalization across multiple subjects, and utilizes mel-frequency cepstral coefficient matrices extracted from multichannel EEG and electrocardiogram (ECG) signals as input features. Trained and validated on the Helsinki neonatal EEG dataset with 10-fold cross-validation, the proposed model achieved an average accuracy of 97.52%, sensitivity of 98.31%, specificity of 96.39%, and F1-score of 97.95%, enabling accurate seizure prediction up to 30 minutes before onset. The inclusion of ECG alongside EEG improved the F1-score by 1.42%, while the incorporation of an attention mechanism yielded an additional 0.5% improvement. To enhance transparency, we incorporated SHapley Additive exPlanations (SHAP) as an explainable artificial intelligence method to interpret the model and provided localization of seizure focus using scalp plots. The overall results demonstrate the model's potential for minimally supervised deployment in neonatal intensive care …

Objective Young children and infants, especially newborns, are highly susceptible to seizures, which, if undetected and untreated, can lead to severe long-term neurological consequences. Early detection typically requires continuous electroencephalography (cEEG) monitoring in hospital settings, involving costly equipment and highly trained specialists. This study presents a low-cost, active dry-contact electrode–based, adjustable electroencephalography (EEG) headset, combined with an explainable deep learning model for seizure detection from reduced-montage EEG, and a multimodal artifact removal algorithm to enhance signal quality. Methods EEG signals were acquired via active electrodes and processed through a custom-designed analog front end for filtering and digitization. The adjustable headset was fabricated using three-dimensional printing and laser cutting to accommodate varying head sizes …

Discrete Fourier Transform (DFT) based beam-formers provide N beams with a uniform linear array having N antennas. Despite low-complexity, DFT beamformers cannot generate deep nulls at arbitrary directions of arrivals (DOAs) in the beam pattern to attenuate interferences. On the other hand, data-dependent adaptive beamformers can achieve deep nulls at arbitrary DOAs, however, at higher complexity due to an inversion of a matrix in deriving optimum weights. In this paper, we propose a combined approximate DFT (ADFT) and adaptive beamformer for a uniform linear array consisting of P subarrays, each consisting of M antennas. Here, we first employ P ADFT beamformers to process signals received by subarrays, and the P outputs of these subarrays are then processed by an adaptive beamformer. Simulation results confirm that the proposed beamformer provides the benefits of both DFT and adaptive …

Objective Neonates are highly susceptible to seizures, which can have severe long-term consequences if undetected and left untreated. Early detection is crucial and typically requires continuous electroencephalography (EEG) monitoring in a hospital setting, which is costly, inconvenient, and requires specialized experts for diagnosis. In this work, we propose a new low-cost active dry-contact electrode-based adjustable EEG headset, a new explainable deep learning model to detect neonatal seizures, and an advanced signal processing algorithm to remove artifacts to address the key aspects that lead to the underdiagnosis of neonatal seizures. Methods EEG signals are acquired through active electrodes and processed using a custom-designed analog front end (AFE) that filters and digitizes the captured EEG signals. The adjustable headset is designed using three-dimensional (3D) printing and laser cutting to …

Five-dimensional (5-D) light field videos (LFVs) capture spatial, angular, and temporal variations in light rays emanating from scenes. This leads to a significantly large amount of data compared to conventional three-dimensional videos, which capture only spatial and temporal variations in light rays. In this paper, we propose an LFV compression technique using low-complexity 5-D approximate discrete cosine transform (ADCT). To further reduce the computational complexity, our algorithm exploits the partial separability of LFV representations. It applies two-dimensional (2-D) ADCT for sub-aperture images of LFV frames with intra-view and inter-view configurations. Furthermore, we apply one-dimensional ADCT to the temporal dimension. We evaluate the performance of the proposed LFV compression technique using several 5-D ADCT algorithms, and the exact 5-D discrete cosine transform (DCT). The experimental results obtained with LFVs confirm that the proposed LFV compression technique provides a more than 250 times reduction in the data size with near-lossless fidelity with a peak-signal-to-noise ratio greater than 40 dB and structural similarity index greater than 0.9. Furthermore, compared to the exact DCT, our algorithms requires approximately 10 times less computational complexity.

Virtual cycling is an engaging way to promote health and fitness while addressing the safety challenges of outdoor cycling. This paper introduces a Competitive Virtual Cycling (CVC) system that enhances user experience through immersive visuals, real-time competition, and low-latency networking. The system integrates virtual reality (VR) headsets, smart bicycle trainers, and real-world 360 ° videos to create a realistic gaming environment. A dynamic resistance feedback mechanism, driven by real-time sensor data from smart trainers and actuators, adjusts pedal resistance based on terrain variations, ensuring an authentic riding experience. To enhance engagement, we propose a novel avatar placement algorithm that optimizes competitor visibility in immersive, interactive races. Performance evaluations show that the system achieves a mean latency of 60 ms on a 4G network and 30 ms on a private 5G network …

Semi-supervised domain generalization (SSDG) leverages a small fraction of labeled data alongside unlabeled data to enhance model generalization. Most of the existing SSDG methods rely on pseudo-labeling (PL) for unlabeled data, often assuming access to domain labels-a privilege not always available. However, domain shifts introduce domain noise, leading to inconsistent PLs that degrade model performance. Methods derived from FixMatch suffer particularly from lower PL accuracy, reducing the effectiveness of unlabeled data. To address this, we tackle the more challenging domain-label agnostic SSDG, where domain labels for unlabeled data are not available during training. First, we propose a feature modulation strategy that enhances class-discriminative features while suppressing domain-specific information. This modulation shifts features toward Similar Average Representations-a modified version …

Robust feature tracking is a basic requisite for indoor robot navigation. However, indoor scenes mostly contain poorly localizable features. As a result, indoor feature tracking without artificial markers is challenging and remains an attractive problem. We propose to solve this problem by constraining the locations of a large number of non-distinctive features by several planar homographies which are computed using distinctive features. Use of a large number of non-distinctive features within the constraints imposed by planar homographies allows us to gain robustness against the absence of artificial landmarks. Also, the lesser computational cost in estimating these non-distinctive features helps to maintain the efficiency of the proposed method. Our outcome is a robust in door feature tracker. In this dissertation, we will establish that multi-homographic in door feature tracking is feasible with the combination of …

Dr. Manjula Siriwardhana is attached to the National Engineering Research and Development Centre of Sri Lanka as a Principal Research Engineer. As a researcher, Dr. Siriwardhana has received postgraduate research scholarships from several countries and conducted original research and published in highly cited journals. She earned her PhD from the Australian National University (ANU), Canberra, ACT, Australia, MESc. from Western University (UWO), London ON, Canada, MEng. In Energy Technology from Asian Institute of Technology (AIT), Thailand and BSc. Eng.(Hons) from the University of Peradeniya, Sri Lanka. She has collaborated on multi-country research projects financed by international organizations including the International Research Institute (IRI) for Climate and Society, START, UNEP, UNDP, GEF, GNESD and EEP Mekong Region. She is experienced in project execution, evaluation …

This research presents a comprehensive experimental and simulation-based investigation aimed at enhancing the link distance and performance of indoor Visible Light Communication (VLC) systems. A basic VLC setup was incrementally improved by optimizing key system parameters, including LED transmit power, transimpedance amplifier (TIA) feedback resistor, lens configuration, photodiode selection, and wall color. Each enhancement was experimentally validated, and corresponding simulations were conducted to analyze system behavior and verify measurement results. The study also examines the applicability of well-established VLC channel models, such as the Masao Nakagawa model and second-reflection models, in real-world scenarios. The findings confirm that incorporating optical and environmental optimizations can significantly extend link distance, up to 3 m with high accuracy, while aligning …

Low-resolution analog-to-digital converters (ADCs) have emerged as a promising solution to reduce the power consumption of massive multiple-input-multiple-output (MIMO) receivers. However, allocating a fixed number of ADC bits uniformly across all receive antennas often results in suboptimal energy efficiency (EE), particularly when channel conditions vary significantly across antennas. This paper proposes a learning-based ADC bit allocation strategy that dynamically adjusts the resolution of each ADC based on the corresponding channel gain. The proposed method effectively allocates higher-resolution ADCs to antennas experiencing stronger channel gains while minimizing the resolution for weaker channels, thereby improving the overall EE. Simulation results demonstrate that the proposed method achieves near-optimal EE performance comparable to the brute-force approach but with significantly lower …

The quadrotor is an underactuated, nonlinear system that presents significant challenges in both modeling and control design. This work develops a decoupled control framework based on the translational (Newtonian) and rotational (Eulerian) dynamics of the quadrotor. A Linear Quadratic Gaussian (LQG) regulator is implemented for control, with two extended Kalman filters employed for state estimation in the respective dynamic subsystems. The full design process, from dynamic modeling to flight simulation presented in detail. Key elements include nonlinear simulation, model linearization, state-space representation, feedforward compensation, Linear Quadratic Regulator (LQR) gain tuning, actuator dynamics, sensor noise, LQG design, and extended Kalman filter. The limitations of applying linear control to a nonlinear system are also presented.

Plasma-assisted processes are critical in semiconductor fabrication, where variations in plasma parameters can lead to reduced yield and device failure. This work presents the design and experimental validation of a non-invasive plasma density sensing system, intended for use in plasma-enhanced process reactors without requiring structural modifications. A prototype sensor array was developed to measure ion current via embedded electrodes and tested in a custom-built low-pressure plasma chamber. Experimental results show a strong linear correlation between chamber pressure and sensor current, confirming the sensor's ability to track plasma density trends. Supporting simulations and data visualization further demonstrate the feasibility of this approach for real-time, non-invasive plasma diagnostics.

Traditional boundary scan chain implementations are typically linear and fixed in length for specific designs, with continuous testing being their primary application. However, these chains also play a critical role in continuous monitoring, where the standard architecture, originally intended for testing, is not fully optimized. In particular, input ports that do not contribute to data capture are still included in the scan path, resulting in unnecessary scanning and reduced efficiency. Since the scan chain is linear, the Test Access Port controller must sequentially access all cells, including irrelevant ones, leading to slower monitoring speeds. In this research, we propose a novel boundary scan architecture that enhances both continuous testing and monitoring by focusing only on relevant input ports for test pattern insertion and output ports for response capture. This targeted approach significantly improves monitoring …

The rapid evolution of Artificial Intelligence (AI) has revolutionized multimedia creation, particularly through Text-to-Image (T2I) diffusion models, which synthesize images from textual descriptions with impressive capabilities. Building on this, enhancing video GIF generation has become a promising frontier. However, the field of Text-to-Video (T2V) GIFs remains underexplored, with limited research addressing this specific application. This synthesis method expands creative expression and finds utility in education, marketing, and entertainment. The lack of focused work highlights both novelty and opportunity for impactful contributions. This study focuses on optimizing T2V GIF generation in terms of computational and parameter efficiency. The resulting model maintains a parameter count below 1 billion, enabling faster training, reduced inference time, lower memory usage, and compatibility with low-end hardware. Inspired …

With the recent advancement of Large Language Models (LLMs), environmental perception methods, and frameworks for simplifying robot functionality implementation, robotic receptionists present an exciting opportunity to minimize humanrelated errors. However, most existing humanoid robots are costly and overqualified for such tasks, as they are primarily designed for research applications. To address this, we introduce a standalone, offline, cost-effective smart mobile robot receptionist built on an edge device that efficiently performs a predefined set of tasks while utilizing minimal hardware and computational resources. Our robot receptionist can engage in smooth conversations, recognize and interact with both new and returning users, perform hand gestures, provide navigation assistance through verbal and physical guidance, and autonomously navigate while avoiding static and dynamic obstacles …

This review outlines the current status and proposes future directions for using auditory brainstem responses (ABRs) to continuous speech as a tool in hearing aid assessment. In current clinical practice, ABRs measured via electroencephalography (EEG) in response to short stimuli, averaged across repetitions, are a cornerstone of objective hearing screening and evaluation. Recent advancements have demonstrated that ABRs can be estimated from EEG recorded during continuous speech listening by using linear models which map the auditory speech features, extracted using auditory models, to continuous EEG activity. This approach enables the simultaneous investigation of multiple auditory pathway stages using complex, ecologically valid sounds; advancing potential future hearing assessments that can be seamlessly integrated into daily life. Current findings indicate that responses similar to click-ABRs …

Cardiac auscultation is an essential clinical skill, requiring excellent hearing to distinguish subtle differences in timing and pitch of heart sounds. However, diagnosing solely from these sounds is often challenging due to interference from surrounding noise, and the information may be limited. Most of the existing solutions that adaptively cancel external noise are either non-real-time or computationally intensive, making them unsuitable for implementation in a portable system. This work proposes an end-to-end system with a real-time adaptive noise cancellation pipeline integrated into a device that simultaneously acquires electrocardiogram (ECG) and phonocardiogram (PCG) signals. We employ a burst adaptive normalized least mean square algorithm that adjusts its adaptation in response to high-energy, non-stationary hospital noise. The algorithm's performance was initially assessed using datasets with artificially induced noise. Subsequently, the complete end-to-end system was validated using real-world hospital recordings captured with the dual-modality device. For PCG and ECG signals recorded from the device in noisy hospital settings, the proposed system achieved signal-to-noise ratio improvements of 37.01 dB and 30.32 dB, respectively. Furthermore, complexity analysis confirms the pipeline's suitability for embedded implementation. These results demonstrate the system's effectiveness in enabling reliable and accessible cardiac screening in noisy hospital environments typical of resource-constrained settings.

Real-time pedestrian sensing at crossings is essential for Intelligent Transportation Systems (ITS). Yet traditional vision-based solutions and wearable devices face privacy, cost, and deployment challenges. While Wi-Fi channel state information (CSI) has shown promise in indoor environments, its potential for dynamic outdoor environments remains largely unexplored. This paper presents a device-free, CSI-based system for detecting human movement and walking direction at crossings and alert oncoming vehicles to enhance pedestrian safety. A twostage classification framework is adopted with noise suppression and Doppler-sensitive features. Classical and deep learning models are evaluated across four outdoor locations and seven subjects, achieving 96.9% accuracy for movement detection and up to 62% accuracy for classification of movement direction. These findings demonstrate the feasibility of …

I am honoured and privileged to welcome you to the Moratuwa Engineering Research Conference 2025 (MERCon 2025), proudly organized by the Engineering Research Unit (ERU) of the University of Moratuwa, Sri Lanka. This year's conference, the 11th edition, is being held on August 14 th and 15 th , 2025, at the University premises. MERCon has evolved over two decades of rich experiences gained through the annual ERU Symposium on Research for Industry and has now grown into a fully-fledged international multidisciplinary engineering conference technically co-sponsored by the IEEE. It continues to showcase innovative and impactful research contributions across a broad spectrum of engineering disciplines.

This study integrates an immersive digital environment to enhance accessibility and engagement with traditional music. Sri Lankan traditional music types such as bullock-cart songs, boatman songs, kamath gee, and watch hut songs were used as musical forms for this study. The proposed VR application provides an immersive educational experience for contemporary audiences. Given the intangible nature of these traditions and their declining practice, the application aimed to bridge the generational gap by leveraging VR technology to recreate these cultural expressions in their original context. The VR environment was designed using a structured design pipeline. Evaluation was conducted to assess the usability, cultural authenticity, and educational effectiveness of the developed VR application. The results demonstrate the potential of the application as an innovative tool for intangible cultural heritage preservation and transmission.

Musical instruments play a pivotal role in the cultural fabric of communities. This research delves into the rich musical heritage of the Tamil community in Batticaloa, Sri Lanka, exploring the diverse range of traditional musical instruments that play a pivotal role in their cultural expressions. Through a comprehensive ethnographic study, this research examine the historical significance, craftsmanship, and sociocultural context surrounding these musical instruments. Data were collected using combination of qualitative interviews, field observations, and archival research to document and analyze the various musical instruments unique to the Tamil peoples of Batticaloa. The Parai, for instance, continues to be associated with the Parai Mela Kooththu, a tradition of dancing with the drum accompanied by a double-reed instrument called Sornali. Other instruments such as Mathalam, Uduku, Nadaswaram, and Thavil each carry its distinct cultural narrative, contributing to the vibrant musical tapestry of the region. Findings reveal not only the cultural significance of these musical instruments but also their multifaceted role in fostering community identity, preserving historical connections, and serving as living embodiments of artistic heritage.

Photoplethysmography (PPG) has emerged as a critical non-invasive technique for cardiovascular health monitoring through wearable devices, yet the physiological mechanisms underlying PPG signal formation are not completely understood. This affects the development of accurate diagnostic tools and personalized health monitoring systems. To overcome it, analysis of how cardiac output and other physiological factors are reflected in PPG, especially at the wrist, is required, and end-to-end PPG simulation methods have been developed for this purpose. However, these methods have limitations in either accuracy or computational efficiency. To address this gap, we present a comprehensive end-to-end framework for simulating wristbased PPG signals by integrating radial artery hemodynamics with optical light transport modeling through skin. In our approach, first, cardiac parameters are used to generate the …