Enhancing IMU-based smartphone context detection using Transformer

Deep learning models like Long Short Term Memory (LSTM) or Convolutional Neural Network (CNN) have yielded great results in the field of understanding smartphone context in recent years. In this paper, we propose using another deep learning model based on Transformer to tackle the same task with IMU sensors data. We will start by reviewing an overview of existing approaches for smartphone context detection. A new sensor-based dataset on smartphone context and step recognition is collected. Extensive experiments have been conducted to compare the accuracy of the Transformer model to other deep learning models like LSTM or CNN. Furthermore, the Transformer model used for smartphone context understanding also improves the efficiency of step counting.