Spike-Driven Neuromorphic Sensing for Energy-Proportional Indoor Air Quality Monitoring in Multi-Zone IoT-Enabled Smart Building Environments

Indoor Air Quality (IAQ) monitoring, especially in multi-zone smart buildings, is typically limited by the high computational and energy requirements of continuous sensor processing, which makes event-driven methods desirable for efficiency. Energy proportionality, in this context, refers to a system whose computational cost scales with the significance of detected environmental changes rather than with the fixed sampling rate. This paper presents a spike-driven neuromorphic sensing framework for decentralized IAQ monitoring that combines adaptive Kalman filter preprocessing, dynamic threshold-based asynchronous spike encoding, and a Leaky Integrate-and-Fire neural network with Spike-Timing-Dependent Plasticity (STDP) learning. Multiple-parameter IAQ data including PM1, PM2.5, PM10, CO2, CO, TVOCs, and O3 were sampled from nine functionally differing zones of an educational building in Metro Manila, Philippines. The neuromorphic model yielded a mean Sparse Firing Ratio of 10.94%, a Mean Response Time of 10.62 timesteps, and an energy efficiency proxy score of 9.28. Neuron population scaling and parameter robustness analyses revealed that the four neurons per parameter were enough to saturate the performance, and FLOP-based estimation indicated an 8.9-fold computational reduction (approximately 89% fewer FLOPs) compared to LSTM inference. In addition, the revised Performance Efficiency Index and composite efficiency score corroborated the stable and energy-proportional nature of behavior in all zones. These results illustrate that spike-based neuromorphic computation is an energy-efficient and scalable way for decentralized smart-building IAQ monitoring, though hardware-level validation on dedicated neuromorphic processors remains necessary for absolute power saving verification. Multi-seed validation (five seeds) with expanded baselines including GRU, Temporal CNN, XGBoost, and Logistic Regression confirmed the robustness and repeatability of reported metrics.