摘要
为了提高人脸识别率,提出了一种优化置信度的判别嵌入式隐马尔可夫(EHMM)人脸识别方法。提出的方法基于假设检验,通过最小化检验错误率得到优化置信度判别式训练准则。在优化置信度判别式训练准则的前提下,通过参数估计求解判别式转换矩阵,提取出具有判别性、低维度的图像特征,确保观察样本能正确地分配到其对应的模型状态,以提高所训练出的EHMM模型的正确识别率。理论分析证明了优化置信度判别式训练准则的有效性,详细的实验及与现有方法的比较结果表明,提出的识别方法具有更好的识别性能。
To improve face recognition rate, this paper proposed a face recognition scheme based on confidence measure discriminative EHMM. Based on hypothesis test, the proposed scheme deduced the optimized confidence measure discriminative training criterion by minimizing mis-verification error rate. With the criterion, the scheme estimated model parameters to obtain discriminative transformation matrix and extract low-dimentional and discriminative image feature, which assured the obsertion samples could be assigned to the corresponding states to improve recognition rate. Theoretical analysis proved the effectiveness of the criterion. Detailed experimental results and comparisons with the existed schemes show that the proposed scheme can get better recognition rate.
出处
《计算机应用研究》
CSCD
北大核心
2010年第5期1987-1990,共4页
Application Research of Computers
关键词
置信度
嵌入式隐马尔可夫模型
人脸识别
confidence measure
embedded hidden Markov model（EHMM）
face recognition
