Evaluation of memory-efficient 1-bit compressed sensing in wireless sensor networks

Wireless sensor networks generate big-data which leads to overwhelming traffic in the networks. Compressed sensing (CS) can exploit signal sparsity to reduce the amount of traffic. We evaluate the performance of 1-bit CS with a circulant random bipolar measurement matrix. The results show that 1-bit CS can decrease traffic at higher degree in wireless sensor networks compared to the conventional CS. Moreover, applying a circulant random bipolar measurement matrix overcomes the limitation of memory in sensor nodes. The scheme is studied in terms of reconstruction accuracy and memory consumption by using real sensor data. The implementation on sensor nodes validates the viability on a practical system.