摘要
基于尺度不变特征变换的特征包(BoF-SIFT)支持向量机的分类方法具有较好的手势识别效果,但是计算复杂度高、实时性较差。为此,提出了融合Hu矩与基于快速鲁棒特征的特征包(BoF-SURF)支持向量机(SVM)的手势识别方法。特征包模型中用快速鲁棒性特征(SURF)算法替换尺度不变特征变换(SIFT)算法提取特征,提高了实时性,并引入Hu矩描述手势全局特征,进一步提高识别率。实验结果表明,算法无论是实时性还是识别率都要高于BoF-SIFT支持向量机方法。
The classification method using bag of features-scale invariant feature transformation （BoF-SIFT） support vector machine got a better result on hand gesture recognition. However, it had a high computational complexity which results in the worse real-time performance. So, this paper proposed the method of combining the Hu moments and bag of features-speeded up robust feature （BoF-SURF） support vector machine. The SURF algorithm replacing the SIFT algorithm extracted the features in the bag-of-features model. It could enhance the real-time. Then, using the Hu moments described the global features of hand gesture image, it could enhance the recognition accuracy. The result shows that the real-time and accuracy of the pro- posed algorithm is higher than BoF-SIFT algorithm.
出处
《计算机应用研究》
CSCD
北大核心
2014年第3期953-956,960,共5页
Application Research of Computers
关键词
手势识别
特征包模型
快速鲁棒特征
HU不变矩
支持向量机
hand gesture recognition
bag-of-features model
speeded up robust feature （SURF）
Hu invariant moments
support vector machine（SVM）
