Coding Projects
Chinese Food Website
A web application showcasing various Chinese dishes, built to be user-friendly and visually appealing.
ASL-to-Speech Interpreter App
Developed using Python and MediaPipe. Converts American Sign Language gestures to audible speech, improving accessibility for the deaf and hard of hearing community.
Multimodal Biometric Authentication Project
Fusion of face and palmprint biometrics using Python, OpenCV, and scikit-learn to enhance authentication security and accuracy, especially when face coverings are present. This system demonstrates improved robustness over single-modality methods by combining the unique features of both traits.
Project Overview
Traditional single-modality biometric systems are vulnerable to spoofing and fail when the biometric input is obscured, a problem highlighted during the COVID-19 pandemic with face masks. This project addresses these issues by fusing facial and palmprint data at the score-level to create a more resilient and accurate authentication system.
Key Research Questions
- How does combining face and palmprint data affect authentication performance compared to a single modality?
- What is the impact of facial occlusions (like masks and glasses) on the multimodal system's performance?
Methods & Datasets
The system was developed using Python, Dlib, and scikit-learn. It leverages two public datasets: the Specs on Faces (SoF) Dataset for facial images with occlusions and the Birjand University Mobile Palmprint Database (BMPD). Features were extracted from each and fed into five classification models (KNN, Decision Tree, SVM, Logistic Regression, LDA).
Key Results
The k-Nearest Neighbors (KNN) classifier performed best for both individual modalities, with an Equal Error Rate (EER) of 0.192 for faces and 0.196 for palmprints. While the face modality showed better separation between genuine and impostor scores, the fusion of both created a more balanced and robust system, validating the hypothesis that multimodal systems are more resilient, especially against occlusions.
Project Documents
The full project report, detailing the methodology, results, and ethical considerations.
A summary of the project findings, as presented to peers and faculty.