Browsing by Author "Piyawardana, V"

Now showing 1 - 7 of 7

Embargo
Adapting MaryTTS for Synthesizing Sinhalese Speech to Communicate with Children
(IEEE, 2021-12-01) Lakmal, M. A. J. A; Methmini, K. A. D. G; Rupasinghe, D. M. H. M; Hettiarachchi, D. I; Piyawardana, V; Senarathna, M; Reyal, S; Pulasinghe, K
The majority of the Sri Lankan population speak Sinhala, which is also the country's mother tongue. Sinhala is a difficult language to learn by children aged between 1–6 years when compared to other languages. Text to speech system is popular among children who have difficulties with reading, especially those who struggle with decoding. By presenting the words auditorily, the child can focus on the meaning of words instead of spending all their brainpower trying to sound out the words. In Sri Lanka, however, computer systems based on the Sinhala language especially for children are extremely rare. In this situation having a Sinhala text-to-speech technology for communicating with children is a helpful option. Intelligibility should be considered deeply in this system because this is specific for children. Recordings of a native Sinhalese speaker were used to synthesize a natural-sounding voice, rather than a robotic voice. This paper proposes an approach of implementing a Sinhalese text-to-speech system for communicating with children using unit selection and HMM -based mechanisms in the MaryTTS framework. Although a work in progress, the intermediate findings have been presented.
Embargo
ASD Screening for Toddlers via Physical Interpretation through Advanced AI
(IEEE, 2021-12-09) Jayasekera, D; Alwis, H; Dissanayaka, H; Mudalinayake, R; Piyawardana, V; Pulasinghe, K
Autism Spectrum Disorders (ASD) are generally causing challenges for significant communication, social interaction, and behavioral patterns to elderly people and children. Providing early treatments can make a huge advancement in the lives of children. Meanwhile, there is a limited number of systems to screen and identify ASD children. This research project is about developing a set of tools bonding together to one system called "AI - Bot Simon" to screen kids with ASD by filling the gap. In the system development process mainly, Audio, Facial expressions, Gestures, and the Gates of a targeted group of children are considered for screening. Since the target group is 6 months to 4 years, they are in early language development age. On the technical side of view Machine Learning (ML) and Deep Learning (DL) with Neural Networks (NN) are used for advanced screening and monitoring for automation of the process. In the last step of the development, all the outputs or information gathered from each tool or model, processed, analyzed, and provided to the users of the system by an Artificial Intelligence (AI) bot implemented with a web application and a mobile application whether children are suffering from ASD or not.
Embargo
Automated Sinhala Speech Emotions Analysis Tool for Autism Children
(IEEE, 2021-08-11) Welarathna, K. T; Kulasekara, V; Pulasinghe, K; Piyawardana, V
— Autism Spectrum Disorder (ASD) is a neurological disorder that impairs children's development and symptoms that can be noticed in early childhood. One of the main diagnosis characteristics of ASD is the child having unusual emotions and expressions during social interactions. The main problem is how to distinguish these symptoms. Only 14 out of 100 Autistic kids, before they reach the age of 24 months, get medical treatments since the unavailability of resources to identify them early. If they can be recognized early, a therapeutic engagement can be done to help them overcome those issues in social interactions, when they reach school-going age. The focus of this research is to develop a tool to screen atypical children from typical children. This research attempts to recognize the correct emotion of a child, while the child is talking. The input audio stream of children was normalized into a specific range, sub-framed into 2s length for language-independent, noise reduction, and age independence features, and extracting the most effective 40 audio features. The Convolutional Neural Network (CNN) based model classifies eight different emotions of sad, disgust, surprise, neutral, happy, calm, fear, and angry with an accuracy matrix of F1 score of 0.90, even in the uncontrol environment. If the classifying emotions have small frequency variances, the trained model has the ability to handle them.
Embargo
E-Pod: E-learning System for Improving Student Engagement in Asynchronous Mode
(IEEE, 2021-10-27) Tennakoon, S; Wickramaarachchi, T; Weerakotuwa, R; Sulochana, P; Karunasena, A; Piyawardana, V
Over the last decade, e-learning has grown significantly as the internet and education have merged to give individuals the possibility to learn new skills. With the COVID-19 pandemic, the use of e-learning has increased in an exponential manner. The asynchronous e-learning mode is found to be appealing to students due to its adoption at any time and in any location. Yet, this mode of learning suffers from lack of interactivity. Under such circumstances, this research proposes E-Pod, an asynchronous e-learning system, which promotes student engagement. Through attention monitoring, when the students are found to be inattentive they are provided with opportunities to engage in a wide range of activities such as summarization activities, puzzles and answering questions to improve the interactivity. The accuracy achieved for the gaze estimation model is 89.5 % and the accuracy achieved for the facial emotion recognition model is 83%. In order to generate FBQ and MCQ questions for students, a SVM model was trained to an accuracy of 95.56%. E-Pod includes a MaLSTM model with 83.98% accuracy for short answer evaluation and a DistilBERT model with 86.8% accuracy for essay answer evaluation. The system is developed using a blend of cutting-edge technologies including image processing, Natural Language Processing, machine learning algorithms and language models. With these features, E-Pod is proposed as an all-inclusive system which stands out from existing e-learning systems and will be helpful for educational institutions to deliver flexible and self-paced learning to their students in asynchronous mode.
Embargo
Individualized Edutainment and Parent Supportive Tool for ADHD Children
(IEEE, 2020-12-10) Thennakoon, A; Perera, D; Sugathapala, S; Weerasingha, S; Samarasinghe, p; Dahanayake, D; Piyawardana, V
Attention-Deficit/Hyperactivity Disorder (ADHD) is a comorbid disorder that can impact a child and his/her family. ADHD children have considerable obstacles in managing time, understanding instructions, and paying attention to the activities. To address these perplexities, this research has designed a mobile application to help parents to have better interaction with the children and for the children to enjoy their learning activities. The specialty of this application is the models are trained on individual child skills and needs. Issues with time management are handled by the Scheduler component while the Instruction Predictor module supports the parent in recognizing the child's understandability level. Furthermore, the children are provided with edutainment activities based on their attention and ability levels. Different models have been used in predicting the results through these modules and the prediction result accuracy exceeds 90% in most of the cases. Out of the many models, The Random Forest model resulted in the best overall performance. The application was tried by many parents and health professionals and received satisfactory and commendable reviews.
Embargo
Non-Verbal Bio-Markers for Automatic Depression Analysis
(IEEE, 2021-12-02) Yashodhika, G. B. O; De Silva, L. S. R.; Chathuranaga, W. W. P K; Yasasmi, D. L. R; Samarasinghe, P; Pandithakoralage, S; Piyawardana, V
Detection of early depression risk is essential to help the affected individual to get timely medical treatment. However, automatic Depression Risk Analysis has not received significant focus in prior studies. This paper aims to propose an Automatic Depression Risk Analyzer based on non-verbal biomarkers; facial and emotional features, head posture, linguistic, mobile utilization, and biometrics. The analysis has shown that facial and emotional features can learn to identify depression risk better when compared with the head pose and emotional features. Moreover, the study shows that Depression Risk Analysis based on linguistic performed well with 95% accuracy for Sinhala content and 96% accuracy for contextual in English. Identifying the depression risk based on the biometrics, the sleep pattern analysis obtained 95% accuracy with the K Nearest Neighbour (KNN). Further, the mobile utilization analysis with the KNN model achieved 81% accuracy towards the Depression Risk Analysis. The accuracy of Depression Risk Analysis can be improved by extending analytic models to work as a single model. Furthermore, The models have been integrated with a mobile application that allows users to get a comprehensive Depression Risk Analysis based on each biomarker. These additional methods will function together to provide a more accurate on assessing depression risk.
Embargo
Standalone Application and Chromium Browser Extension-based System for Online Examination Cheating Detection
(IEEE, 2021-12-09) Kariyawasam, S; Lakshan, A; Liyanage, A; Gimhana, K; Piyawardana, V; Mallawarachchi, Y
Educational organizations and institutes that provide services to the public use e-learning frequently than before. The incapacity to evaluate the knowledge acquired is a flaw in education. Due to the current situation, traditional evaluation and examinations are not possible. In a developing country like Sri Lanka, the conduct of online examinations has not been efficient, resulting in cheating at examinations due to vulnerabilities resulting from organizational policies and the difficulty to track down candidates who are prone to cheating, therefore use of facial features for candidate verification and to monitor the background interactions the use of audio and video is taken into consideration with the aid of two cameras; the system mounted camera and a wearable camera containing a microphone allowing audio detection. In this research, we suggest using the training data set generated from individuals to undertake a training approach to improve the robustness for background interactions through audio and video to detect the level of cheating of candidates.