Publications
2022
Job Search during COVID-19: How Online Mindfulness Intervention helped to Reduce Stress and Enhance the Self-esteem of the Job Seeker
Authors: Rajalakshmi S A, Sowndaram Sowndaram, Preetham Ganesh, Harsha Vardhini Vasu
This study examines the impact of mindfulness-based intervention (MBI) on stress and self-esteem among college students involved in campus placements. Using a sample of 100 students from Amrita Vishwa Vidyapeetham University, evaluations with stress and self-esteem scales showed that MBI significantly reduced stress and increased self-esteem. The findings suggest policymakers should promote MBI in curricula and fund training initiatives to support students during challenging transitions.
2021
POS-Tagging based Neural Machine Translation System for European Languages using Transformers
Authors: Preetham Ganesh, Bharat S. Rawal, Alexander Peter, Andi Giri
This study addresses language barriers by proposing a novel Neural Machine Translation (NMT) approach using inter-language word similarity and Part-of-Speech (POS) tagging for model training and testing. Two classical architectures, Luong Attention-based Sequence-to-Sequence and Transformer models, were used, with tokenization by SentencePiece and Subword Text Encoder, respectively. The models were evaluated on Spanish, French, and German datasets with BLEU, Precision, and METEOR scores, showing promising results.
2021
Continuous American Sign Language Translation with English Speech Synthesis using Encoder-Decoder Approach
Authors: Preetham Ganesh
This thesis explores bridging communication barriers for the deaf and hard-of-hearing by converting ASL videos into English speech using deep learning. The process involves four steps: recognizing ASL phrases, converting them to English text, translating text to phonemes, and generating spectrograms. A Sentence-based ASL dataset was created, and models were developed for each step, using Seq2Seq and Transformer architectures with datasets like WLASL, ASLG-L12, CMUDict, and LJSpeech to achieve accurate ASL-to-English translation.
2020
Personalized system for human gym activity recognition using an RGB camera
Authors: Preetham Ganesh, Reza Etemadi Idgahi, Chinmaya Basavanahally Venkatesh, Ashwin Ramesh Babu, Maria Kyrarini
This paper presents a Human Activity Recognition system using an RGB camera to classify gym activities (e.g., push-up, squat) through models like SVM, Decision Tree, KNN, and Random Forest, with the latter achieving 98.98% accuracy. A repetition counter was developed using local minima analysis and dynamic time warping to assess workout accuracy per skeletal point. An interactive Android app was also built to provide users insights into their workouts.
2020
Estimation of Rainfall Quantity using Hybrid Ensemble Regression
Authors: Preetham Ganesh, Harsha Vardhini Vasu, Dayanand Vinod
This study tackles rainfall prediction in Tamil Nadu, India, using ensemble methods like bagging and boosting with optimized parameters. Bagging Regression achieved the best results, though differences among models were minimal. Various ensemble techniques, including Simple Averaging, Blending, and Stacking, were applied to combine models, and performance was evaluated through graphical analysis against actual rainfall values.
2020
Forecast of Rainfall Quantity and its Variation using Environmental Features
Authors: Preetham Ganesh, Harsha Vardhini Vasu, Dayanand Vinod
This paper develops three models—District-Specific, Cluster-Based, and Generic-Regression—to predict monthly rainfall across districts in Tamil Nadu, India, with a focus on capturing sudden rainfall fluctuations. Each model is tailored to different geographic and climatic groupings, allowing district-wise comparisons to identify the most accurate model. Additionally, the study examines monthly rainfall variations across regions.
2020
Juxtaposition on Classifiers in Modeling Hepatitis Diagnosis Data
Authors: Preetham Ganesh, Harsha Vardhini Vasu, Keerthanna Govindarajan Santhakumar, Raakheshsubhash Arumuga Rajan, Bindu K R
This study compares classification models—SVM, Random Forest, Decision Tree, Logistic Regression, and Naive Bayes—on Hepatitis C data, evaluating performance metrics like accuracy, precision, and recall. Using Z-score normalization and an 80-20 train-test split, Random Forest achieved the highest accuracy at 90.7%, outperforming the other models. The analysis was conducted in R.