Human Motion Recognition Based on Acceleration Characteristics

In recent years, the pedometer device has been used more and more widely in smart phones. In this paper, an algorithm for human motion recognition and detection is designed according to the acceleration transducer built in the mobile phone. According to the user’s habit of placing the phone in the jacket pocket (there is no restriction on the direction of the phone), we can read out the data of the three-axis acceleration sensor in the exercise mode, this article sets five human body movement modes: walking mode, running mode, upstairs mode, downstairs mode, and stationary state. The collected data are processed and multiple sets of features are extracted for principal component analysis. Then the dimensionality-reduced feature data set is distinguished and identified as data input. After analyzing and comparing the algorithm, support vector machine (SVM) is used to classify and model. In the process of classification, parameters of SVM are optimized in order to improve classification accuracy.