Hand gesture recognition of static letters American sign language (ASL)‎ using deep learning

Abstract EN

An American Sign Language (ASL) is a complex language.

It is depending on the special gesture stander of marks.

These marks are represented by hands with assistance by facial expression and body posture.

ASL is the main communication language of deaf and people who have hard hearing from North America and other partsof the world.

In this paper, Gesture recognition is proposed of static ASL using Deep Learning.

The contribution consists of two solutions to the problem.

The first one is resizedwith Bicubic static ASL binary images.

Besides that, good recognition results in of detection the boundary hand using the Robert edge detection method.

The second solution is to classify the 24 alphabetsstatic characters of ASL using Convolution Neural Network (CNN) and Deep Learning.

The classification accuracy equals to 99.3 % and the error of loss function is 0.0002.

According to 36 minutes with 15 seconds of elapsed time result and 100 iterations.

The training is fast and gives the very good results, in comparison with other related works of CNN, SVM, and ANN for is a complex language.

It is depending on the special gesture stander of marks.

These marks are represented by hands with assistance by facial expression and body posture.

ASL is the main communication language of deaf and people who have hard hearing from North America and other partsof the world.

In this paper, Gesture recognition is proposed of static ASL using Deep Learning.

The contribution consists of two solutions to the problem.

The first one is resizedwith Bicubic static ASL binary images.

Besides that, good recognition results in of detection the boundary hand using the Robert edge detection method.

The second solution is to classify the 24 alphabetsstatic characters of ASL using Convolution Neural Network (CNN) and Deep Learning.

The classification accuracy equals to 99.3 % and the error of loss function is 0.0002.

According to 36 minutes with 15 seconds of elapsed time result and 100 iterations.

The training is fast and gives the very good results, in comparison with other related works of CNN, SVM, and ANN for training.