摘要
利用FRFT的时频双重特性和LBP算子能提取纹理图像微小特征的优点,提出一种将分数阶Fourier变换(FRFT)与局域二值模式(LBP)算子相结合的笑脸识别算法。对训练样本进行分数阶Fourier变换,取其变换的幅值信息作为脸部表情特征,与LBP融合进行分类判别,同时采用总体识别率和笑脸识别率统计结果,在RML表情数据库进行仿真验证。实验结果表明,该方法在笑脸识别中相比其他方法的识别性能更好。
For time-frequency characteristics of Fractional Fourier Transform(FRFT) and Local Binary Pattern(LBP) operator have advantages in extracting texture feature,this paper proposes to combine FRFT with LBP for recognizing smile emotion.It takes the amplitude information of transformed FRFT as facial feature,fuses LBP and discriminates smile with proper classification criterion.At the same time,the overall recognition rate and smiling face recognition rate are used to express experimental results.Compared with other methods,simulation results in RML expression database show that the method is effective for smile recognition.
出处
《计算机工程》
CAS
CSCD
2012年第20期169-171,175,共4页
Computer Engineering
基金
国家自然科学基金资助项目(61071211)