Sign Language Recognition Using Wearable Electronics: Implementing k-Nearest Neighbors with Dynamic Time Warping and...

Area of Science:

• Computer Science
• Biomedical Engineering
• Artificial Intelligence

Background:

• Sign language recognition is crucial for communication accessibility.
• Existing systems often lack comprehensive movement tracking.
• Wearable technology offers a promising avenue for real-time sign language interpretation.

Purpose of the Study:

• To develop and evaluate a sign language recognition system using wearable electronics.
• To compare the performance of k-Nearest Neighbors with Dynamic Time Warping and Convolutional Neural Networks for sign language classification.
• To assess the system's accuracy in recognizing a set of Italian and international sign words.

Main Methods:

• A wearable system comprising a sensory glove and inertial measurement units was developed to capture hand, wrist, and arm movements.
• Two classification algorithms were implemented: k-Nearest Neighbors with Dynamic Time Warping and Convolutional Neural Networks.
• Seven participants (five male, two female) performed 100 repetitions of ten distinct sign words, including Italian and international terms.

Main Results:

• The k-Nearest Neighbors with Dynamic Time Warping classifier achieved an accuracy of 96.6% ± 3.4%.
• The Convolutional Neural Networks classifier demonstrated a higher accuracy of 98.0% ± 2.0%.
• Both classifiers exhibited high performance, outperforming many existing sign language recognition systems.

Conclusions:

• The proposed wearable sign language recognition system is effective and highly accurate.
• Convolutional Neural Networks offer superior performance compared to k-Nearest Neighbors with Dynamic Time Warping for this task.
• The system represents a significant advancement in wearable technology for sign language interpretation.