Research On Service Robot Interaction Based On Expression And Voice

With the development of science and technology,the rise and wide application of artificial intelligence,service robots can gradually replace some human labor and complete more thoughtful service work.To carry out efficient service work,the effectiveness of interaction is essential.How to make service robots provide better services to users through natural humancomputer interaction has become a research hotspot.Expressions and voices convey almost all the information in the process of people’s communication.Voice is the main channel of people’s daily communication,and facial expressions contain more than half of the emotional information.This paper firstly studies the related technologies of facial expression recognition and expression,voice interaction based on the application requirements of service robots,and then designs and implements the interaction of service robots based on facial expressions and voices.The main work is as follows:(1)Through the selection and improvement of lightweight convolution,a lightweight expression recognition system suitable for service robot interaction is designed and built.Aiming at the problems that the system has high dependence on the training set,low recognition rate and poor generalization performance in the real environment,the designed lightweight expression recognition system is optimized from the perspective of the training data set.The optimization focuses on the noise adjustment of the dataset and designs a method for applying belief learning.This method corrects the self-confidence,and adjusts the noise detection range by introducing two sets of hyperparameters to perform certain manual intervention on the confidence learning results.Experiments show that the recognition performance of the recognition system after the training set noise adjustment has been significantly improved.(2)Supplement research materials related to facial expression recognition and voice interaction for the elderly in special groups with great potential in the service robot market.Research the elderly facial expression data set,design a standardized sample processing and labeling method,design and implement the elderly data set production software based on this,and realize the data set generation;on-the-spot visits to collect the elderly voice intent sample set,based on which to realize the objection Expansion of gallery intent,regular corpus and special corpus,and based on this sample set,the recognition effect of the researched speech intent recognition method is tested.(3)A voice-based interaction scheme is designed and implemented.A relatively complete intent database and an intent recognition strategy suitable for channel fusion and special expressions are designed.In view of the problem that some special language expressions cannot be captured and recognized and fed back when users generate demands in voice interaction,the special intent corpus is expanded,and the Combined with the designed template matching recognition method,these special corpora can be accurately recognized in the unified standard recognition framework.(4)An expression-based interaction scheme is designed according to application requirements.Aiming at the problem that the user’s facial expression information is not fully utilized,a real-time emotion monitoring mechanism and a facial intention feedback mechanism are designed;at the same time,in order to increase the interaction temperature,a dynamic cartoon image is designed for real-time interaction of expressions.(5)The service robot fusion interaction scheme is designed and implemented,and the conflict problem of each channel is solved.Based on this,a service robot interaction system is designed to achieve effective interaction.