| |
| Human motion recognition is becoming a research upsurge, which aims at understanding human behavior, and plays an increasingly important role in a number of applications, such as health care and smart home. In this paper, we collect datasets by using the built-in sensors of a mobile phone and propose an approach to extract features based on wavelet transform. In contrast to the existing related works, our work intends to recognize the physical activities when the phone's orientation and position are varying. The activities' true acceleration is inferred by using the phone's pitch, yaw and roll angles. After preprocessing, the continuous original time series data is segmented into discrete training samples by the sliding windows of proper size. Then statistical features such as wavelet coefficients are extracted through the wavelet transform. Support Vector Machine (SVM) is employed as classifier to recognize five types of motion: jumping, walking, running, stepping upstairs and stepping downstairs. We find a proper wavelet basis function to extract the features and achieve an average recognition accuracy of 90.71%. We can distinguish the five kinds of motion clearly, so the results show that it is feasible to use wavelet transform to extract features in human motion recognition. |
|
| Keywords:Human motion recognition; Wavelet transform; Support vector machines (SVM) |
|
