DeepWave: A Wavelet-Based Technique for Compressing In-Field Sensors Data

The real-time use of wireless sensor networks has become increasingly popular due to their attractive features. However, energy management becomes a critical factor in these networks, especially during deployment, because each sensor node has limited battery capacity. While wireless sensors offer advantages, transmitting massive volumes from numerous sensors to a central data center wirelessly presents a significant hurdle. In this work, the focus is on the seismic sensor (geophone), and the challenge lies in transmitting hundreds of recordings per geophone through narrowband channels without overloading the data center or the sensors themselves. This motivates our proposed method, DeepWave, a lightweight and standalone compressive sensing approach designed specifically for in-field data acquisition. This study presents an effective method for compressing seismic data in the field, followed by the utilization of the wavelet transform and integration with a convolutional neural network (CNN) for recovery at a later stage. The sparsity-aware schematic, DeepWave, is proposed for compressed data recovery and compared with benchmarking techniques. A key strength of this method is that it is general and works with any underlying data statistics, allowing it to adapt to a wide range of exploration and sensing scenarios. Our findings indicate that this CNN-based approach achieves an effective balance between data compression (93.75% compression percentage) and signal fidelity ( dB normalized mean-square error) on the evaluated dataset.