Saliency based alphabet and numbers of American sign language recognition using linear feature extraction
Majid Zamani, Hamidreza Rashidy Kanan · 2014
Sign language is a way to communicate for deaf people, which hand shapes are used instead of sound patterns. In this paper, we present a method for recognizing alphabet and numbers in American sign language based on saliency detection of the image. After detecting saliency, the images were processed by Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA), in order to reduce dimensions and maximize class internal similarity and minimize class external similarity. Finally resulting vectors will be trained and will be classified by the neural networks (NN). The use of this system is communication with deaf people in addition to connecting with computer; this is due to the use of standard letters in sign language. Also experiments were carried out on a new standard dataset in this field. The recognition rate of the system was 99.88% using 4-fold cross validation method in 4 training terms on average. The results of the proposed system represent high accuracy and proper performance of this method compared with the others.