Research On Robot Intelligent Voice Interaction System Based On ROS

In recent years,with the rapid development of the field of robotics and natural language processing,the research on robot intelligent voice interaction technology has also entered a stage of rapid development.If a robot wants to further exert its own intelligence,it must have super human-computer interaction capabilities.Among the many human-computer interaction methods,voice-based human-computer interaction is the most important and convenient interaction method.The main research content of this paper has the following aspects.Firstly,in view of the complex and low accuracy of the current intention recognition process,a user intention recognition classification model based on Glove＿Bi GRU＿Self-attention is proposed to recognize the user’s intention.In this model,the Bi GRU network combines the advantages of the bidirectional cyclic neural network and the long-and short-term memory network,which can more fully learn the relationship between contexts.In addition,the calculation of GRU tensor is not much,so the training speed is faster.In order to capture more direct semantic dependencies,this paper inputs the fully connected output of the Bi GRU model into the Attention mechanism layer,and uses the Attention mechanism to encode user questions.So as to improve the accuracy of intention recognition,the feasibility of this model is proved through experiments.Secondly,in response to the situation where there is no network or the network quality is not good,the robot voice interaction function cannot be used normally,a robot intelligent voice interaction system with both online and offline voice interaction functions is designed.Then the above-mentioned intention recognition classification model is applied to the system,and at the same time,the voice interaction system is divided into four modules: speech wake-up,speech recognition,intention recognition,and speech synthesis.On the basis of designing and implementing each module separately,the modules are integrated through ROS,and finally the entire robot intelligent voice interaction system is formed.Finally,the feasibility of the system is proved by verifying the system in terms of recognition rate and recognition time.And on the basis of the realization of the entire intelligent voice interaction system,the voice interaction function and the robot movement function are integrated to achieve the purpose of controlling the movement of the robot through voice.