Abstract Face recognition task has been an active research area in recent years in computer vision and biometrics. Feature extraction and classification are the most significant steps for accurate face recognition systems. Conventionally, the Eigenface approach or frequency domain features have been used for feature extraction, but they are not invariant to outdoor conditions like lighting, pose, expression, and occlusion. Multiple convolutional and pooling layers of Deep Learning Networks (DLN) will efficiently extract the face database’s high-level features in the present work. These features have given to the Kernel Extreme Learning Machine (KELM) classifier, whose parameters have optimized using Particle Swarm Optimization (PSO). The proposed Deep Convolutional-Optimized Kernel Extreme Learning Machine (DC-OKELM) algorithm leads to better performance results and fast learning speed than classification using deep neural networks. The performance of DC-OKELM has evaluated on four standards face databases: AT&T, CMU-PIE, Yale Faces, and UMIST. Experimental results have compared with other state-of-the-art classifiers in terms of error rate and network training time, which shows the proposed DC-OKELM classifier’s effectiveness.

中文翻译：

---

深度卷积 - 基于优化内核极限学习机的人脸识别分类器

摘要 人脸识别任务近年来一直是计算机视觉和生物识别领域的一个活跃研究领域。特征提取和分类是准确人脸识别系统最重要的步骤。传统上，特征脸方法或频域特征已被用于特征提取，但它们对光照、姿势、表情和遮挡等室外条件并非一成不变。在目前的工作中，深度学习网络 (DLN) 的多个卷积和池化层将有效地提取人脸数据库的高级特征。这些特征赋予了内核极限学习机 (KELM) 分类器，其参数已使用粒子群优化 (PSO) 进行了优化。所提出的深度卷积优化内核极限学习机 (DC-OKELM) 算法比使用深度神经网络的分类具有更好的性能结果和更快的学习速度。DC-OKELM 的性能在四个标准人脸数据库上进行了评估：AT&T、CMU-PIE、Yale Faces 和 UMIST。实验结果在错误率和网络训练时间方面与其他最先进的分类器进行了比较，这表明了所提出的 DC-OKELM 分类器的有效性。