Design And Implementation Of Human Machine Speech Interaction System Based On ROS

With the development of artificial intelligence,more and more companies and institutions engage in robot research,and various types and functions of robots have emerged.In the future,the robot will gradually take on some simple,repetitive or heavy work and become people's daily life assistant,or even become a member of the family.Speech,as a convenient,fast and efficient way,will undoubtedly be an important and indispensable communication method between human and robot,so it is necessary to design a human machine speech interaction system.The existing human machine speech interaction systems have poor functions.Some system acquires the answer only by matching people's question with the local question sets.If people's question does not belong to the local sets,the system can only give a general reply.The system usually has a poor generality and scalability.Some system is based on the online interface and cannot be used normally without the Internet.We design a human machine interaction system based on Robot Operating System(ROS)in order to solve the problem the current interaction system have.ROS is simple to use and has a good architecture,which has being used by more and more companies and institutions on robot research field.D esigning human machine speech interaction system based on the ROS being modular at a fine-grained scale concept can make the research and debugging of the various modules easily.In this thesis,we mainly focus on designing and implementation of human machine speech interaction system based on ROS.First,in order to fulfill the human machine speech interaction function,we build the question libraries and the robot control command library separately.We propose the idea that the question libraries should be divided into the daily frequently asked questions(FAQ)library,the professional FAQ library and the real-time questions library,which can make the management and maintenance of each library convenient.Meanwhile we can also customize the robot by replacing the professional FAQ library.Second,we adopt the Bayesian classification algorithm to process the text information received from speech recognition,and determine whether the text information is about the daily life,the professional field,the real-time questions,or the robot control commands.Then we can apply the sentence similarity calculation method to acquire the similarity value between the text information and the questions of the question library or the commands of the robot control command library.According to the maximum similarity value,we can determine what the robot should do.The classification of text information not only reduces the number of sentence similarity calculation,but also improves the response speed of the robot.Third,we analyze the functional modules that the human machine speech interaction system should have.We design the communication way between the modules based on ROS,define the information format of the communication based on ROS,and give the process of whole system in detail.Finally,in order to verify the performance of our proposed human machine speech interaction system based on ROS,we test the human machine speech question-answering function and the voice control function on PC and robot platform respectively.The experiment results verified that the proposed human machine speech interaction system based on ROS has a strong reliability and practicality.