The Improvement of Behavior Recognition Accuracy of Micro Inertial Accelerometer by Secondary Recognition Algorithm
Yu Liu, Le Wang, Yang Cao, Zhen Fang, Yi Ou · 2014
Behaviors of still, walking, running, jumping, upstairs and downstairs can be recognized by micro inertial accelerometer of low cost. By using the features as inputs to the well-trained BP artificial neural network which is selected as classifier, those behaviors can be recognized. But the experimental results show that the recognition accuracy is not satisfactory. This paper presents secondary recognition algorithm and combine it with BP artificial neural network to improving the recognition accuracy. The Algorithm is verified by the Android mobile platform, and the recognition accuracy can be improved more than 8 %. Through extensive testing statistic analysis, the recognition accuracy can reach 95 % through BP artificial neural network and the secondary recognition, which is a reasonable good result from practical point of view. Copyright © 2014 IFSA Publishing, S. L.