标题：Semantic recognition of signed language using convolutional neural network
作者机构：[Wang, L] School of Foreign Languages and Literature, Shandong University, No. 27, Shanda Nanlu, Jinan, 250100, China
通讯作者地址：[Wang, L] School of Foreign Languages and Literature, Shandong University, No. 27, Shanda Nanlu, China;
来源：ICIC Express Letters, Part B: Applications
关键词：Convolutional neural network; Gesture recognition; Landmark Localization; Signed language
摘要：In this paper we study the hand gesture recognition problem for signed language. The complex background in real world application is a major challenge for hand region segmentation. First, we adopt the skin color model to pre-process the input image. Second, skin color filtered images are used to train a fast convolutional neural network for hand region detection. The detection is converted into a regression problem which leads to more efficient detection results. Third, the detected hand image patch is applied with a state-of-the-art landmark localization algorithm using Markov Random Fields and Active Shape Models. Finally, the American signed language is recognized in this paper for verifying the proposed system. Experimental results show that compared with several other approaches, the proposed gesture recognition system achieves satisfactory performance. © 2017 ICIC International.