A Visual-based Research On Static Sign Language Recognition

Sign language recognition research was started in20th century,80years. Itinvolves pattern recognition, image processing, computer science and so on. Signlanguage is a parallel language with natural language which is used by deaf people indaily life to communicate. The purpose of sign language recognition is to facilitatecommunication between normal people and the deaf. It can help deaf people integrateinto social life more conveniently. So sign language recognition research has animportant application value. In the same time, it uses the latest scientific andtheoretical result, so it has important theoretical value.As hands are elastic objects and the same kind of sign gesture has big differences,so it will generate shadows in collection gestures. And the non-smooth hand surfacewill generate shadows too. All these factors bring difficulty to the research system.This paper focuses on gesture area extraction from complex background, gesturelibrary establishment, image preprocessing, feature extraction and recognition.First a template library was established in the lab. Then the imgaes weresegmented by the skin color clustering characteristic in an improved YCbCr colorspace. This method eliminates the step of graying, and the segmentation results arebetter than the method of graying. Then uses the method of smoothing, morphologyand region grow method to enhance the gesture images. The second part of the paperintroduces feature extraction. The system extracts the continuous outer contours of thebinary image. Then choose the Fourier descriptors and Hu Moments as theeigenvectors. Finally, the paper uses Hausdorff distance as the standard to classify andrecognize the sign language gestures.The experimental results show that the algorithm employed in this article has ahigh computing speed and recognition rate and some theoretical innovation and socialvalues.