Sign Language Recognition for Deaf People Using Novel HMM Compared with CNN and LSTM

The aim of this study is to enhance sign language recognition for the deaf by implementing a HMM and improving its accuracy with the use of Convolutional Neural Networks (CNN) and Long Short-Term Memory Networks (LSTM). The HMM was applied to a dataset to accurately recognize sign language letters. This study compares the performance of the HMM model with a CNN+LSTM model to assess which provides better accuracy in sign language recognition. The research involved a sample size of <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\mathbf{1 0 0}$</tex> participants per group, with a power value set at 0.8. The results indicated that the HMM model achieved a remarkable accuracy of 99 %, significantly outperforming the CNN+LSTM model, which achieved an accuracy of 90 %. The findings suggest that the HMM model not only offers superior accuracy but also a lower mean error compared to the CNN+LSTM model. This research demonstrates that the HMM approach is more effective for recognizing sign language, providing a valuable tool for improving communication for the deaf community.