For energy-limited applications of electronic nose, an application-specific architecture is essential to realize a low-energy gas recognition system. In this paper, a pseudo-zero-leakage gas recognition system is proposed to recognize different gases using event-driven power control. Additionally, this gas recognition system can recognize four different gases with concentration information by drift-insensitive on-line training, achieving 100% recognition accuracy for gas type and 89.4% accuracy for concentration analysis. For further reducing the overall energy consumption, both near-threshold SRAM and low-voltage embedded ReRAM are integrated into the proposed system, respectively. Based on TSMC 65nm LP CMOS process, the total energy of the gas recognition systems with SRAM and ReRAM are only 8.62μJ and 2.04μJ in a sensing period, respectively. Hence, an energy-efficient gas recognition system can be realized by a pseudo-zero-leakage event-driven structure with ReRAM.