“…As deep learning techniques have developed, researchers have begun applying deep learning methods to various motion or global positioning system (GPS) sensors. Studies have applied convolutional neural networks (CNNs), ensemble CNNs, convolutional long shortterm memory (convolutional LSTM), or other self-designed architectures to the features extracted from GPS sensors, and they achieved up to 92.7% accuracy [9], [10], [20], [23]. In contrast, studies using motion sensors have applied neural networks (NNs), deep neural networks (DNNs), CNNs, recurrent neural networks (RNNs), or other self-designed architectures, resulting in accuracies up to 98.4% [3], [4], [6], [8], [10], [24], [26].…”