In the realm of human-computer interaction, the quest for more accessible communication methods has driven the development of Hand Gesture Recognition Systems. This project aims to contribute to this advancement by creating a detection system capable of recognizing hand forms through a combination of carefully crafted preprocessing algorithms and the utilization of Convolutional Neural Networks (CNNs) for image classification.
The experimental methodology employed in this project focuses on the American Sign Language (ASL) Alphabet dataset. The following steps outline our approach:
- Import and Resize Images: Standardize images by resizing.
- Convert to Grayscale: Extract feature intensity values through grayscale conversion.
- Gaussian Noise Removal: Enhance reliability by smoothing edges through noise reduction.
- Otsu’s Thresholding and Inverted Binary Thresholding: Analyze intensity distribution for effective image segmentation.
- Canny Edge Detection: Identify hand boundaries in the frame.
- CNN for Gesture Classification: Pair gestures using a Convolutional Neural Network.
- Evaluate Metrics: Assess model performance using metrics such as Accuracy, Precision, F1 Score, ROC Curve, and Confusion Matrix.
To illustrate our methodology, consider the ASL Alphabet dataset, where each gesture represents a distinct sign in American Sign Language. The algorithm processes these gestures, identifies hand forms, and utilizes CNN to classify and pair them accurately.
- ASL Alphabet Dataset - Image dataset for alphabets in the American Sign Language.
A Brief Technical Report