This paper addresses an autonomous facial expression recognition system using the feature selection approach of the Quantum-Inspired Binary Gravitational Search Algorithm (QIBGSA). The detection of facial features completely depends upon the selection of precise features. The concept of QIBGSA is a modified binary version of the gravitational search algorithm by mimicking the properties of quantum mechanics. The QIBGSA approach reduces the computation cost for the initial extracted feature set using the hybrid approach of Local binary patterns with Gabor filter method. The proposed automated system is a sequential system with experimentation on the image-based dataset of Karolinska Directed Emotional Faces (KDEF) containing human faces with seven different emotions and different yaw angles. The experiments are performed to find out the optimal emotions using the feature selection approach of QIBGSA and classification using a deep convolutional neural network for robust and efficient facial expression recognition. Also, the effect of variations in the yaw angle (front to half side view) on facial expression recognition is studied. The results of the proposed system for the KDEF dataset are determined in three different cases of frontal view, half side view, and combined frontal and half side view images. The system efficacy is analyzed in terms of recognition rate.

中文翻译：

---

受量子启发的二元引力搜索算法识别面部表情

本文介绍了一种使用受量子启发的二元引力搜索算法 (QIBGSA) 的特征选择方法的自主面部表情识别系统。面部特征的检测完全依赖于精确特征的选择。QIBGSA 的概念是通过模仿量子力学的特性，对引力搜索算法进行了修改的二进制版本。QIBGSA 方法使用局部二进制模式与 Gabor 滤波器方法的混合方法降低了初始提取特征集的计算成本。所提出的自动化系统是一个顺序系统，对基于图像的 Karolinska 定向情绪面孔 (KDEF) 数据集进行了实验，该数据集包含具有七种不同情绪和不同偏航角的人脸。进行实验以使用 QIBGSA 的特征选择方法和使用深度卷积神经网络进行分类以进行稳健和有效的面部表情识别来找出最佳情绪。此外，研究了偏航角（正面到半侧视图）变化对面部表情识别的影响。所提出的 KDEF 数据集系统的结果是在正面视图、半侧视图以及组合正面和半侧视图图像这三种不同情况下确定的。从识别率方面分析系统效能。研究了偏航角变化（从正面到半侧视图）对面部表情识别的影响。所提出的 KDEF 数据集系统的结果是在正面视图、半侧视图以及组合正面和半侧视图图像这三种不同情况下确定的。从识别率方面分析系统效能。研究了偏航角变化（从正面到半侧视图）对面部表情识别的影响。所提出的 KDEF 数据集系统的结果是在正面视图、半侧视图以及组合正面和半侧视图图像这三种不同情况下确定的。从识别率方面分析系统效能。