| dc.contributor.advisor | Sign Language is the main communication
medium among deaf and speech impaired people. In order
to express their thoughts and emotions, hand gestures are
used. In Sri Lanka, the Sri Lankan Sign Language is
considered as the native Sign Language. But unfortunately,
in the Sri Lankan community, the deaf and speech
impaired people are often ignored by society due to the
language barrier. As a solution to the problem, this paper
proposes a system to assist the deaf and speech impaired
people in capturing their sign-based message via the
camera and then convert it into Sinhala text and
furthermore into audio form. So, the main aim of this
research is to eliminate the communication gap and to
improve interaction between them and the common
people. Convolutional Neural Networks (CNN) has been
used as the technology of this research. The proposed CNN
model which consists of one convolution layer, one max
pooling layer and two dense layers along with Relu and
Softmax activation functions has the ability to
automatically extract the features of the input static
gesture and recognize it (out of 24 classes) and give it as
the output in text form. And then a text to speech engine
will eventually generate the audio output in the Sinhala
language. The model was trained for more than 20 times
and obtained an accuracy of 98.61%. The proposed model
has been implemented through python and libraries like
OpenCV, Keras, pickle, etc have been used in advance.
Keywords— Convolutional Neural Networks, Static
gestures, Gesture recognition, HSV | |
