(英) |
In this paper, we first transform the 3-axis acceleration data of smart phone, which is left in our laboratory, from the device coordinate system into world coordinate system. Based on the transformed acceleration, we use various deep learnings, LSTM, BiLSTM, GRU, BiGRU, and CNN-LSTM for user identification. As a result, we found that the acceleration data of the previous research can be correctly transformed by the extended Kalman filter. We also confirmed that the user identification using deep learning can achieve a high identification rate even for the 3-axis acceleration after the coordinate transformation. |