基于Gabor特征的遗传算法实现静态手势识别

Gabor-based Genetic Algorithm for Static Gesture Recognition

在人机交互领域,手势识别是一个重要的研究分支。为了提高手势识别的准确率,提出一种基于Gabor特征结合遗传算法的手势识别方法。数据集采用网络数据源,将该数据源中的图片分成一个小样本数据库和大样本数据库两部分,使用小样本数据库进行训练测试后,再使用大样本手势数据库进行实验结果测试。针对小样本数据集,数据库图像原来是在RGB空间,为便于提取图像特征,将这些图像转换至LAB空间,之后使用Gabor特征提取技术来提取每张图片中的纹理特征,通过该技术来获得手势特征,通过调整参数使用遗传算法训练数据,最后选用大样本数据库的数据库中的10种数字手势识别图像进行验证。结果表明,对给定数据库的数字手势进行识别检验,该方法具有较好的鲁棒性。

In the field of human-computer interaction, gesture recognition is an important research branch. In order to improve the accuracy of gesture recognition, a gesture recognition method based on Gabor feature and combined with genetic algorithm is proposed. The data set adopts the network data source, and the picture in the data source is divided into two parts: a small sample database and a large sample database. After the training test with small sample database, the large sample gesture database is used to test the experimental results. For small sample data sets, the database image is originally in RGB space, in order to facilitate the extraction of image features, these images are converted to LAB space, and then the texture features in each image are extracted by Gabor feature extraction technology, and via this technique, the gesture features obtained, and the genetic algorithm is used to train the data by adjusting the parameters. Finally, 10 kinds of digital gesture recognition images in the database of the large sample database are selected for verification. The recognition test of digital gestures in a given database shows that this method has fairly good robustness.