计算机工程与应用
計算機工程與應用
계산궤공정여응용
COMPUTER ENGINEERING AND APPLICATIONS
2014年
12期
137-143
,共7页
三维人脸识别%超平面学习%余弦相似性度量%支持向量机%线性判别边信息
三維人臉識彆%超平麵學習%餘絃相似性度量%支持嚮量機%線性判彆邊信息
삼유인검식별%초평면학습%여현상사성도량%지지향량궤%선성판별변신식
three-dimensional face recognition%hyperplane learning%cosine similarity measure%Support Vector Machine (SVM)%side-information based on linear discriminant
针对三维人脸识别中受光照、姿态、表情等变化而影响识别性能的问题,提出了一种原型超平面学习算法。利用SVM将弱标记数据集中的每个样本表示为一个原型超平面中层特征,使用学习组合系数从未标记的通用数据集中选择支持向量稀疏集;借助于Fisher准则最大化未标记数据集的判别能力,使用迭代优化算法求解目标函数;利用SILD进行特征提取,余弦相似性度量完成最终的人脸识别。在USCD/Honda、FRGC v2、LFW及自己搜集的人脸数据集上的实验结果表明,该算法优于其他几种三维人脸识别算法。
針對三維人臉識彆中受光照、姿態、錶情等變化而影響識彆性能的問題,提齣瞭一種原型超平麵學習算法。利用SVM將弱標記數據集中的每箇樣本錶示為一箇原型超平麵中層特徵,使用學習組閤繫數從未標記的通用數據集中選擇支持嚮量稀疏集;藉助于Fisher準則最大化未標記數據集的判彆能力,使用迭代優化算法求解目標函數;利用SILD進行特徵提取,餘絃相似性度量完成最終的人臉識彆。在USCD/Honda、FRGC v2、LFW及自己搜集的人臉數據集上的實驗結果錶明,該算法優于其他幾種三維人臉識彆算法。
침대삼유인검식별중수광조、자태、표정등변화이영향식별성능적문제,제출료일충원형초평면학습산법。이용SVM장약표기수거집중적매개양본표시위일개원형초평면중층특정,사용학습조합계수종미표기적통용수거집중선택지지향량희소집;차조우Fisher준칙최대화미표기수거집적판별능력,사용질대우화산법구해목표함수;이용SILD진행특정제취,여현상사성도량완성최종적인검식별。재USCD/Honda、FRGC v2、LFW급자기수집적인검수거집상적실험결과표명,해산법우우기타궤충삼유인검식별산법。
The performance of traditional three-dimensional face recognition is impacted by variations of illumination, poses and expression, for which a prototype hyperplanes learning algorithm is proposed. SVM sparse set is selected from generic data set without labeled by learning combination coefficient. Fisher discriminative criterion is used to maximize discriminat ability under the constraint of combination sparse coefficient of SVM model, and the objective function is solved by iterative optimization algorithm. SILD is used to extract features and cosine similarity measure is used to finish face recognition. Experimental results on FRGC v2, USCD/Honda, LFW and a database searched show that proposed algo-rithm has better recognition efficiency than several other existing algorithms.