一种基于改进的SURF算法的静态手语字母识别方法

A novel static sign language letter recognition method based on improved SURF algorithm

结合Kinect传感器提出一种改进的SURF(speeded up robust features)算法进行静态手语字母识别的方法。Kinect传感器采集深度图像进行手势分割可以克服光照变化、复杂背景带来的干扰;改进的SURF算法对实时图像与模板图像的积分图进行计算分析,提取两者的SURF关键点描述符,采用最近邻匹配算法对SURF算法自有的快速索引匹配的结果进行优化,克服了角度旋转变化对手语字母识别率的影响。实验证明,该方法在应对光照变化、复杂背景、角度旋转方面有很好的鲁棒性,平均识别率为97.7%。

In this paper,a novel method of sign language letter recognition based on improved SURF algorithm is proposed,which is combined with Kinect sensor.The Kinect sensor can overcome interference caused by illumination changes and complex background when collecting deep image and fulfill hand gesture segmentation.Integral image of the real-time image and the template image is calculated and analyzed by the improved SURF algorithm,both SURF interest point descriptor are extracted.Then fast indexing algorithm matching result of SURF is optimized by nearest neighbor matching algorithm,which can conquer the influence of angle variation.The experiments show that the method has good robustness in illumination,complex background and angle,and achieves an average recognition rate of 97.7%.