Danish Sign Language Translation

Overview

This project aims to translate Danish Sign Language into text by analyzing video data. We explore and compare the effectiveness of different data extraction techniques, specifically MediaPipe and Unity-VR, and various machine learning models to determine the most accurate setup for sign language translation.

Key Features

• Data Extraction: Utilizes MediaPipe and Unity-VR for landmark representation from video data.
• Machine Learning Models: Employs models including k-nearest neighbors, Random Forest, and neural networks (RNNs and LSTMs) for sign language translation.
• Data Smoothing: Applies low-pass filters and interpolation for data preprocessing.
• Dimensionality Reduction: Utilizes PCA and UMAP for efficient data analysis.

Dataset

The dataset comprises professional-grade videos collected from tegnsprog.dk and matcen.dk, along with additional videos recorded to cover a subset of 9 signs from Danish Sign Language.