Novel Wearable System to Recognize Sign Language in Real Time

Area of Science:

• Biomedical Engineering
• Human-Computer Interaction
• Machine Learning

Background:

• Existing sign language recognition systems often face limitations such as high costs, susceptibility to environmental factors, and complex hardware.
• There is a need for practical, accurate, and user-friendly solutions to bridge communication gaps for the hearing impaired.

Purpose of the Study:

• To develop a practical, real-time sign language word recognition software utilizing surface electromyography and inertial measurement unit data from both arms.
• To overcome the limitations of existing systems by offering a cost-effective and robust solution.

Main Methods:

• Development of a software solution integrating digital signal processing and machine learning algorithms.
• Creation of a dataset comprising 80 frequently used sign language words.
• Feature extraction and performance evaluation using various classification algorithms.

Main Results:

• The Random Forest algorithm achieved the highest recognition accuracy at 99.875%.
• The Naïve Bayes algorithm yielded the lowest accuracy at 87.625%.
• The proposed system demonstrates high efficacy in recognizing sign language words.

Conclusions:

• The developed software offers a practical and highly accurate solution for real-time sign language recognition.
• This system has the potential to significantly improve communication accessibility for individuals with hearing disabilities.
• The system ensures seamless integration into daily life without compromising user comfort or lifestyle.