Design And Implementation Of Smart Home Mobile Robo

Applying mobile robots to the home environment is a development trend of smart homes.Embedded voice recognition,image processing and motion control are the core technologies of home mobile robots.In this study,a mobile robot based on smart home is designed,which can perform speech recognition on simple vocabulary and motion control based on skeleton tracking.The hardware architecture adopts a hierarchical design,including a motion control layer,an information analysis decision layer and a human-computer interaction layer.They are respectively responsible for the implementation of motion control strategies,skeleton tracking and making corresponding motion control decisions and speech recognition.The coupling between the layers is low and the reusability is high.In the design of speech recognition function,TMS320C5509A is selected as the main processor to build an end-to-end speech signal detection device,which can realize the functions of speech signal preprocessing,feature extraction,clustering of syllables by finals,etc.The Gaussian mixture hidden Markov model performs matching recognition and recognizes voice information.In the design of the motion navigation function,the Kinect vision sensor and skeleton tracking algorithm are used to obtain the target skeleton position data,and the linear velocity and angular velocity of the robot are evaluated based on these position data to assist the robot to track the target movement.In the MatLab simulation experiment of speech recognition related algorithm performance,the experiment was carried out by acquiring the vocabulary sound wave data entered by the DSP test experimenter.The results show that the cutting method based on waveform cross-correlation can effectively separate the vowels.The evaluation of the pitch test found that the corresponding recognition range of the speech recognition system proposed in this study is[130Hz,320Hz],and it is determined that the pitch frequency at 230Hz can achieve the best clustering effect The GMM-HMM model based on divided subsets has a recognition rate of more than 95%for simple words,which saves about 23%of the time compared to the original model.In the testing of skeleton tracking and motion control,the left and right hand gestures and the multi-person layout were selected as application scenarios.The results show that the robot can judge the target signal and can display the rough outline of the target to be tracked according to the signal strength;Test the motion control of the robot in different situations,including narrow mobile environment,multi-person movement interference,and target loss.The results show that the robot can adjust the speed according to the environment,have good anti-interference ability,and can recover the lost target to a certain extent.The above results indicate that the home mobile robot designed by this institute can recognize simple control instructions,accurately identify and track targets,and has practicality,which provides a strong guarantee for the application of intelligent robots in the home environment.