Movement Classification based on Acceleration Spectrogram with Dynamic Time Warping Method

In this paper, we propose a movement classification model using acceleration spectrogram combined with dynamic time warping method. First, the proposed system collects data sets from a mobile device embedded acceleration sensor, and filters only coordinate data, which has an effect on the states of movements. Then the data sets in the form of frequency are converted into spectrograms. The RGB color vectors in these spectrograms are used to classify the states of movements. In order to classify these changes, a DTW algorithm is used. Finally, we validate the feasibility and applicability of the proposed model by implementing and applying it to three typical human navigational movements, namely, plain walking, going upstairs and going downstairs.