期刊名称:Journal of Theoretical and Applied Information Technology
印刷版ISSN:1992-8645
电子版ISSN:1817-3195
出版年度:2016
卷号:88
期号:3
出版社:Journal of Theoretical and Applied
摘要:The objective is to generate a basis for sign language recognizer under simple backgrounds. Complications arise in extracting shapes of hands and head using traditional segmentation models due to non-uniform lighting. This paper proposes a wavelet based fusion of two weak edge detection models. One is morphological subtraction model and the other is gradient based canny edge operator. Elliptical Fourier descriptors provide shape models with optimized number of shape descriptors. Principle components determined keep the feature vector to a minimum to accommodate all the frames in the video sequence. Classification of the signs is achieved by training a neural network trained with back propagation algorithm. The proposed method is exclusively tested many times with different examples for correct recognition sequence. Finally, the recognition rate stands at 92.34% when compared to similar model using discrete cosine transform based features at 81.48%.