Compressive sensing (CS) is a novel technique to realize the low-power designs for sensor nodes and reduce overall transmission power in a wireless sensor network. The reconstruction engines using CS… Click to show full abstract
Compressive sensing (CS) is a novel technique to realize the low-power designs for sensor nodes and reduce overall transmission power in a wireless sensor network. The reconstruction engines using CS techniques have also been widely explored to realize real-time processing. However, for the applications of physiological signal monitoring, we are also concerned about physiological conditions. Moreover, in most cases, we are more interested in those high-risk signals, such as paroxysmal atrial fibrillation (AF), that the syndrome happened occasionally. Therefore, lots of computational efforts are wasted if we fully reconstruct those normal signals that are irrelevant to diseases. In this article, we present a tri-mode CS-based compressed analytics (CA) engine that is fabricated in 40-nm CMOS technology and this engine can realize CA, on-demand reconstruction, and full reconstruction in a hardware sharing manner. With CA, we can classify high-risk signals directly in the compressed domain. While the on-demand reconstruction can avoid unnecessary energy consumption of reconstructing these normal sinus rhythm (NSR) signals. Hence, by adopting on-demand reconstruction, the area-energy efficiency (AEE) of this engine can be 3.22-to-67.57 $\times $ better compared with state-of-the-art designs. In summary, the proposed 2.41-mm2 tri-mode CA engine has more comprehensive functionalities, but is more lightweight for medical telemonitoring applications.
               
Click one of the above tabs to view related content.