作者: Neeru Rathee
DOI: 10.1109/CCAA.2016.7813748
关键词:
摘要: Visual Speech processing is the key concern of researchers working in field speech and computer vision. Though earlier audio was popularly used for recognition but their performance deteriorated presence noise. Moreover, variation accent another challenge that affects such systems. In presented paper, we explore lip texture features visemes. The variations temporal behavior coded using Local Binary Pattern three orthogonal planes. classification carried out back propagation neural network, which a network with hidden layer. added advantage layer reading it takes into account nonlinear while speaking. proposed approach Hindi word achieved high accuracy at cost computation time.