As intelligent sensing and smartphone technologies have progressed, companies collect smartphone data from the app and upload it to servers to provide better daily services. Over time, a large amount of data has been accumulated, placing a heavy burden on the storage of smartphones and servers. Data compression is one strategy to solve this problem.Compressed sensing is one data compression method that significantly reduces compression time compared to traditional methods. However, traditional compression sensing methods are time-consuming and have limited reconstruction performance.
On the other hand, deep compressed sensing has recently become popular in computer vision(CV) field. This research aims at compressing sensor data via deep compressed sensing. Firstly, we collected daily sensor data, including accelerated speed, gyroscope, location, etc., and then designed a compressed sensing model based on the Convolutional Neural Network (CNN) and compressed the sensor data. We plan to improve the model by network structure and object function and conduct large-scale data collection experiments in the future. Ultimately, we aim to compress sensor data automatically in real-time as the smartphone collects the data.