Industrial Internet of Things (IIoT) has been widely applied in smart home, smart city, smart traffic, etc. It is a big challenge to recognize facial expression of IIoT systems more effectively. Current facial recognition methods only utilize singular facial images, so accurate features that are highly correlated with facial changes can hardly be extracted. In order to overcome this difficulty and improve facial expression recognition, in this article, we propose a parallel neural network combining texture features, which can be applied in facial expression recognition. This parallel neural network is constructed by convolution neural network, residual network, and capsule network. Additionally, texture analysis is conducted on facial expression images to extract abundant features. Eight texture features are extracted by gray-level co-occurrence matrix and integrated with features of original images. Finally, these integrated features extracted by three kinds of networks are used to classify facial images. Experimental results prove that the proposed approach has a high recognition rate and strong robustness compared to competitive algorithms. Remarkably, our accuracy reaches 98.14%, with an increase of 3.71% in comparison with ResNet, and the F1-score of 0.9801. It is thus verified from this result that the proposed algorithm has many outstanding advantages. The idea in combination with texture features also provides a new solution for image classification.

中文翻译：

---

并行神经网络结合纹理特征的工业物联网面部表情识别


工业物联网（IIoT）已广泛应用于智能家居、智慧城市、智能交通等领域，如何更有效地识别IIoT系统的面部表情是一个巨大的挑战。目前的面部识别方法仅利用单一的面部图像，因此很难提取与面部变化高度相关的准确特征。为了克服这一困难并提高面部表情识别，在本文中，我们提出了一种结合纹理特征的并行神经网络，可应用于面部表情识别。该并行神经网络由卷积神经网络、残差网络和胶囊网络构建。此外，还对面部表情图像进行纹理分析，提取丰富的特征。通过灰度共生矩阵提取八个纹理特征，并与原始图像的特征相结合。最后，利用三种网络提取的综合特征对面部图像进行分类。实验结果证明，与竞争算法相比，该方法具有较高的识别率和较强的鲁棒性。值得注意的是，我们的准确率达到了 98.14%，与 ResNet 相比提高了 3.71%，F1 得分为 0.9801。由此结果验证了所提出的算法具有许多突出的优点。这种与纹理特征相结合的思想也为图像分类提供了一种新的解决方案。