Research On Multi-angle Facial Expression Recognition For Counter Service

As one of the most intuitive ways for people to express their emotions and intentions,facial expressions are widely used in medical,security and driving fields.Among the recognition difficulties caused by the large number of parameters for the recognition of facial expressions by deep learning that are difficult to deploy to embedded platforms and multi-angle facial deflection,this paper studies the expression recognition algorithm under multi-facial posture,and designs a multi-angle expression monitoring system for counter service on this basis,so as to supervise and improve the quality of tel ler service and improve customer business experience.To solve the problem of large number of parameters in conventional deep learning,a lightweight facial expression recognition algorithm is proposed in this paper.The MTCNN face detection algorithm and face alignment technology are used to perform image preprocessing,which can improve the model’s ability to extract important information about facial expressions.Based on the densely connected convolutional networks model,the standard convolution is improved to a depth separation convolution,which reduces the computational effort and the parameters for network training.Based on this,a convolutional attention mechanism module was added to focus on important facial features,and finally realize lightweight and efficient expression recognition.Experimental data show that the number of parameters on the FER2013 dataset is only 0.16 M,and the recognition rate can reach 78.82%,but the expression recognition rate under facial deflection still needs to be improved.In order to improve the lightweight model to cope with the problem of low multi-angle expression recognition rate,the generative adversarial networks(GAN)face correction model is selected and improved.Dual channel mode to ensure image clarity and improve the generalisation capability of the mesh.Improved mesh structure of generators and discriminators to improve the ability to actively display models on faces.On the basis of retaining the original local discriminator,two local discriminators are added to strengthen the discrimination ability of local features of eyebrows,mouth and nose,and facilitate the retention of the model’s representational features.The improved GAN model and lightweight expression recognition model are combined to form the multi-angle expression recognition algorithm proposed in this paper.The experimental results show that the recognition rate of the proposed multi-pie face recognition algorithm is 93% in the KDEF and CMU multi-pie face datasets,and the recognition rate for real scenes is also above 90%.To address the problems of low efficiency and lack of standardised statistical methods in traditional quality management for window services,this paper recognizes the expressions of customers when handling business based on the proposed multi-angle expression recognition model and feeds them back to the background for data processing,and displays the results on the interface to realize the statistical supervision of the service quality of tellers,and finally realiz es the intelligent expression monitoring system for counter services.