Flexible and Self-Adaptive Sense-and-Compress for Sub-MicroWatt Always-on Sensory Recording

Abstract

Miniaturized sensory systems for IoT applications experience a severe power burden from their wireless link and/or embedded storage system. Compressive sensing techniques target data compression before storage and transmission to save power, while minimizing information loss. This work proposes a self-adaptive sense-and-compress system, which consumes only 45-884n W while continuously recording and compressing signals with a bandwidth up to 5kHz. The flexible system uses a combination of off-line Evolutionary Algorithms, and on-line self-adaptivity to constantly adapt to the incoming sensory data statistics, and the current application quality requirements. The 0.27mm <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> sense-and-compress interface is integrated in a 65nm CMOS technology, together with an on-board temperature sensor, or can interface with any external sensor. The scalable, self-adaptive system is moreover heavily optimized for low-power and low-leakage, resulting in a tiny, efficient, yet flexible interface allowing always-on sensory monitoring, while consuming 2.5X less power compared to the current State-of-the-Art.