The Research And Implementation Of Technology In Speech Interaction Of Service Robot

With the rapid development of artificial intelligence technology and intelligent city,service robot gradually comes into people's life and is widely used in various industries such as catering,household,and logistics,which can help humans to efficiently accomplish some tasks repetitive,complex or entertaining.This makes people have higher requirements for a good human-robot interaction experience.However,traditional robot control commonly relies on specialized pointing devices and specific operation flows,which not only consume a lot of resources but also are hard for ordinary users,especially for the special community.During practical applications,such operation requirements are often very difficult to set up or bring an awful user experience.As a kind of natural interaction,speech interaction is the most natural and convenient way for human beings to obtain external information.Compared with other interactive ways,speech interaction makes the robot "listen and speak" and "think and wonder" like a human being.It is of great significance for people to get rid of the physical devices such as traditional mouse,keyboard and touch screen by strengthening the research of speech interaction technology,to enhance the user experience and practicality of human-robot interaction.Based on the research results of speech interaction technology,this paper focuses on the design of service robot's speech interaction system,and attempts to improve system performance by combining compressive sensing technology.The main work includes the following two aspects:First of all,this paper studies the technology of intelligent speech assistant,analyzes the software structure of Robot Operation System(ROS)and the characteristics of Android system,and designs the functional modules of speech recognition,speech synthesis,voice wakeup,semantic understanding and robot control respectively.It finally put forward a speech interaction system based on the cross-platform communication in embedded environment,and it also builds a corpus for the embedded environment to improve the efficiency of robot control.In the real environment of mobile and humanoid,after repeated testing,service robots achieve good experimental results,which satisfies the requirements of embedded robot transplantation platform and the user experience of intelligent house environment.Aiming at reducing the huge resource depletion caused by long-term voice interaction in the voice monitoring system,while embedded devices will have limited storage space and computing resources when using multiple peripherals,it secondly proposes a speech recognition model based on compressed sensing observations which can replace the speech recognition module in voice interaction system to ensure the efficiency and real-time of voice interaction.This paper fully validates the recognition accuracy and real-time performance of the proposed model for speech commands during robot control,and the experiments show that the recognition rate can reach 95% in noisy environments.