Summary
In recent years, with mature motion recognition field motion capture technology and the rapid development, because without using any conventional input device of the computer system can accurately identify the user's intention, the operation of the three-dimensional data processing and data mining techniques, has been widely to use the computer animation, games, new human-computer interaction and intelligent home control and other fields.
SVM (Support Vector Machine, SVM), with its advantages in small training samples, nonlinear and high dimensional pattern recognition and wide attention. In this paper, the study of classical SVM binary classification algorithm, on this basis will SVM algorithm is extended to a number of categories. Further acceleration sensor data acquired by smart phone, a gyroscope and an orientation sensor, built a movement of data collection, transmission and storage platforms, supports multi-user transmission data stored in its action. SVM operation using data pretreatment multiple classification algorithm training, and using particle swarm optimization (PSO) algorithm to optimize the parameters of SVM, the act of establishing classification model, experiments show that the model can accurately identify the rate of 97.30% of the user's intended action.
In order to verify the use scene of a motion recognition system based on an SVM classifier, the article will be applied to the smart appliance control field of home, the software to build an intelligent home simulation module can simulate a series of status information entities of the intelligent home (e.g., turn on the light) . Classification learning operation by the user data, can achieve the purpose of controlling appliance functions like a switch operation by the operation status information, to find an application scenario for the action recognition system based on SVM classifier.
[Key words] action recognition SVM multi-classification PSO smart home
