Real-time Ukrainian sign language recognition system

This article focuses on the development of computational methods and software for the computerized real-time Ukrainian sign language recognition system. The proposed recognition model differs from known approaches in the identification of hand shape in motion. The system makes use of a video camera as a sensor. Hand shape recognition method based on fingertips location and pseudo 2-dimentional image deformation model is used. The sign identification method is based on the hidden Markov models. There has been applied the one-dimensional hidden Markov model which consists of 8 states. Achieved 91.7% sign recognition rate of test set comprising 85 signs.