CAJ | 학술논문

针对光照变化人脸识别问题中传统的光谱回归算法不能很好地进行特征提取而严重影响识别性能的问题，提出了局部判别嵌入优化光谱回归分类的人脸识别算法。计算出训练样本的特征向量；借助于数据的近邻和分类关系，利用局部判别嵌入算法构建分类问题所需的嵌入，同时学习每种分类的子流形所需的嵌入；利用光谱回归分类算法计算投影矩阵，并利用最近邻分类器完成人脸的识别。在两大人脸数据库扩展YaleB及CMU PIE上的实验验证了该算法的有效性，实验结果表明，相比其他光谱回归算法，该算法取得了更高的识别率、更好的工作特性，并且降低了计算复杂度。
침대광조변화인검식별문제중전통적광보회귀산법불능흔호지진행특정제취이엄중영향식별성능적문제，제출료국부판별감입우화광보회귀분류적인검식별산법。계산출훈련양본적특정향량；차조우수거적근린화분류관계，이용국부판별감입산법구건분류문제소수적감입，동시학습매충분류적자류형소수적감입；이용광보회귀분류산법계산투영구진，병이용최근린분류기완성인검적식별。재량대인검수거고확전YaleB급CMU PIE상적실험험증료해산법적유효성，실험결과표명，상비기타광보회귀산법，해산법취득료경고적식별솔、경호적공작특성，병차강저료계산복잡도。
The performance of face recognition with illumination variation is impacted seriously by using traditional spec-tral regression algorithms to extract features, so an algorithm of spectral regression classification optimized by local dis-criminative embedding is proposed. Feature vectors of training samples are calculated. Local discriminative embedding is used to construct embedding needed by classification and embeddings needed by sub-manifold of each classification is learned based on neighbor and classification relationship. Spectral regression classification algorithm is used to compute project metrics, nearest neighbor classifier is used to finish face recognition. The effectiveness and robustness of proposed algorithm has been verified by experiments on the two common face databases extended YaleB and CMU PIE. Experimental results show that proposed algorithm has higher recognition accuracy, better operating characteristic and simpler calculate complexity clearly than several other spectral regression algorithms.