Detecting Sign language gesture for Deaf and Mute using InceptionV3 MODEL
ID:118
View Protection:ATTENDEE
Updated Time:2026-07-22 16:10:08 Hits:16
Online
Abstract
Effective communication is essential for social inclusion, yet individuals with hearing or speech impairments often face communication barriers when sign language is not understood by the wider community. In this study, an automated sign language gesture recognition model was developed to translate static hand gesture images into written text. The Sign Language MNIST dataset was used, which contains 27,455 training images and 7,172 testing images across 24 American Sign Language alphabet classes. Each grayscale image was converted into a three channel RGB representation, resized to 299×299 pixels and normalized before classification. Transfer learning was applied using the InceptionV3 architecture with ImageNet pretrained weights and the model was trained using the Adam optimizer with a learning rate of 0.0001. Performance was evaluated using accuracy, precision, recall, F1-score and a confusion matrix. The proposed model achieved 99% training accuracy and 100% accuracy on the official testing split. These results indicate that InceptionV3 can extract discriminative visual features from static hand gesture images. Nevertheless, validation on more diverse real world images and continuous video streams is needed before the system can be considered ready for practical use.
Keywords
deaf and mute,neural network,InceptionV3,sign language gestures,Bistable Stochastic Resonance; Wavelet Transform; Image Processing,deep learning
Submission Author
HANAN HALAWANI
Najran University
Baseldbwan Baseldbwan
ALBAHA PRIVATE COOLEGE OF SCIENCE
Submit Comment