An Energy-Efficient Multimode Multichannel Gas-Sensor System With Learning-Based Optimization and Self-Calibration Schemes

This paper presents an energy-efficient intelligent multi-sensor system for hazardous gases, whose performance can be adaptively optimized through a multi-mode structure and a learning-based pattern recognition algorithm. The multi-mode operation provides control capability on trade-off relationship of accuracy and power consumption. In-house micro-electro-mechanical (MEMS) devices with a suspended nanowire structure are manufactured to provide desired characteristics of small size, low power, and high sensitivity. The pattern recognition to combine the dimensionality reduction and the neural network is adopted to improve the selectivity of MEMS gas sensors. Moreover, potential deviations in sensing characteristics are calibrated through a proposed self-calibration zooming structure. Reconfigurable circuits for these key features are integrated into an adaptive readout integrated circuit (ROIC) which is fabricated in a 180-nm complementary metal-oxide semiconductor (CMOS) process. For its system-level verification, a wireless multi-channel gas-sensor system prototype is implemented and experimentally verified to achieve 2.6 times efficiency improvement.