Aiming at the problem of single product functions in the domestic pet supplies market,this paper designs a device that can identify the three motion states of running,jumping and flipping.Using the open-source elec-tronic platform Arduino,gyroscope and accelerometer sensors are used to collect motion state data of cats,and the deep learning framework TensorFlow is used for training.It compresses model parameters through Tiny Machine Learning(TinyML)technology to deploy the model on the development board.Finally,by using the trial and error method to find the model structure with the best training effect,the accuracy rate of identifying the action state of cats can ultimately reached more than 90%.
关键词
微型机器学习/运动状态识别/模型搭建/模型训练
Key words
TinyML/Action state identification/Model building/Model training
维普
万方数据