Research And Implementation Of Tibetan Speech Control Of ROS Ground Mobile Robot

In recent years,all kinds of intelligent robots have sprung up,they are changing our production and our lives constantly.More and more companies and research institutions are investing a lot of human and financial resources in the research of this field,and related laboratories have been established in universities,too.Voice is the most direct,simple and efficient way of expression for human communication,the richest carrier of information transfer,and it is also an important way for intelligent robots to interact with human beings.Tibetan is an important member of the Chinese ethnic family,and Tibetan is the most widely spoken minority language in the country.With the socio-economic development of the Tibetan region,it becomes increasingly urgent for Tibetan region people to use intelligent technologies and products.But as the economic,social and technological development of Tibetan region is lagging behind the rest of the country,there are no officially released intelligent robot products for Tibetan language,and no robot research team has been established.Based on the development needs of the Tibetan region,in this thesis,Tibetan speech control for ground mobile robots depended on the Robot Operating System(ROS)is studied and realized,including the study and realization of Tibetan language speech recognition and the design and realization of motion control for ground mobile robots.For Tibetan speech recognition,this thesis studies two speech recognition methods based on dynamic time planning(DTW)algorithm and Gaussian mixture hidden Markov(GMM-HMM)model.First,combined with the actual application needs of robot control,100 control commands(49 in Lhasa and 51 in Amdo)were collected,and a corpus of 5000 voice files was created.Secondly,1000 corpora are selected as the test set,and the remaining 4,000 corpora are used as the training set.Finally,the accidental template training method is used for model training in DTWbased speech recognition,and the method's speech recognition rate is verified to be 90.9%;the model training is performed in the speech recognition based on Gaussian mixed hidden Markov model,and the The speech recognition rate of this method is 94.4%.In the other part,ROS ground mobile robot control is studies.the thesis focuses on the robot operating system ROS1.0 architecture features,engineering structure and communication mechanism,and finally designs and implements a mobile robot control system based on DTW and GMMHMM Tibetan speech recognition.The system was tested on the Turtle Bot2 mobile robot platform.Tests show that the system has good usability,and the recognition rate of Tibetan voice commands is high,and the ground mobile robot moves smoothly after receiving the commands.