Research On Man-machine Interaction Of Mobile Robot Based On Speech Recognition

At present,robots have entered production and life,and a lot of repetitive labor has been replaced by machines.At present,image-based robot control system has been widely used,but as the most convenient and natural way of human-computer interaction,"voice" is still limited.Compared with industrial robots,service robots in the home have an especially urgent need for voice.In this paper,a human-computer interaction system based on speech processing,natural language processing and task analysis is designed for human-computer interaction of service robots in the home.Related research work is as follows:Aiming at the man-machine interaction problem of mobile robot in speech recognition,this paper proposes a method to process commands from the perspective of user and robot.The goal is to get the robot to understand as many commands as possible without affecting the user's language habits.The specific solution of this concept consists of four subsystems: speech processing,natural language processing,task decomposition and simulation system.Aiming at the modules of speech wake up,speech recognition and speech synthesis in speech processing,this project is developed on the basis of commercial algorithm.The main advantage is that under the premise of satisfying the needs of speech processing,it is relatively simple to get a usable system that can be redeveloped.Aiming at the natural language processing problem in mobile robot human-machine interaction,this paper proposes a method combining morphology and syntax.Specifically,the text is processed into linear results by lexical analysis first,and then the linear results of lexical analysis are processed into tree results and the key information is extracted.Lexical problems is essentially a sequence labeling,instructions for the sentence pattern,in this paper is deep learning approach model is more complicated to use the information in the context of sequence annotation,annotation ability but need a lot of data set for training,according to the actual situation of this topic research content of the sentence by conditions random field method to deal with.Syntactic problem is a prediction problem of nonlinear structure,which is to dig deeper information of language on the basis of lexical analysis.In this paper,the linear result obtained from the previous step is processed into a tree structure that can represent the inter-word dependency,and the key information extracted from the tree result is extracted to obtain the core relationship in the sentence.In order to meet the needs of mobile robot task decomposition.This paper proposes a method to apply answer set programming to Chinese instruction task analysis.In this method,the results extracted from key information are converted into predicates in answer set programming by formulating predicate transformation rules,and the user's abstract instructions are processed into atomic instructions that can be executed by the robot by using iclingo solver.Aiming at the problem of execution verification,a visual simulation interface is designed in this paper to simulate and verify the robot's execution of user instructions.Specifically,QT,a cross-platform C++ development framework,is adopted to set up the indoor environment,layout the scene and control the movement of the mobile robot.The simulation system can verify the correctness of the instruction processing of this topic and visually observe the movement of the robot.