针对传统的分类方法由于提取的特征比较单一或者分类器结构过于简单, 导致手语识别率较低的问题, 本文将深度卷积神经网络架构作为分类器与多特征融合算法进行结合, 通过使用纹理特征结合形状特征做到有效识别。首先纹理特征通过LBP、卷积神经网络和灰度共生矩阵方法得到, 其中形状特征向量由Hu氏不变量和傅里叶级数组成。为了避免过拟合现象, 使用“dropout”方法训练深度卷积神经网络。这种基于深度卷积神经网络的多特征融合的手语识别方法, 在“hand”数据库中, 对32种势的识别率为97.73%。相比一般的手语识别方法, 此方法鲁棒性更强, 并且识别率更高。

To solve the problem that the traditional classification method has a relatively single extraction feature or the structure of the classifier is too simple, leading to a low recognition rate of the sign language, in this paper, the deep convolutional neural network architecture was combined as a classifier and a multi-feature fusion algorithm, and it was effectively identified by using texture features and shape features. Firstly, the texture features were obtained by LBP, convolutional neural network and gray-level co-occurrence matrix methods, in which the shape feature vector consisted of Hus invariant and Fourier descriptors. To avoid overfitting, the "dropout" method was used to train deep convolutional neural networks. This kind of multi-feature fusion sign language recognition method based on deep convolutional neural network has a recognition rate of 97.73% for 32 gestures in the “hand” database. Compared with the general sign language recognition method, this method is more robust and has a higher recognition rate.