A Low-Cost IoT Framework for Device-Free Human Counting and Lighting Energy Control in Commercial Spaces

Human activities significantly influence building energy consumption, with optimal electrical appliance utilization relying on accurate real-time occupancy detection. However, existing methods, such as physical sensors, wearable devices, and image recognition techniques, face challenges including extensive deployment complexity and privacy concerns. Unlike traditional automatic control systems, this study introduces BI-Tech, a novel low-cost IoT framework combines device-free human counting with a proactive user reminder mechanism, designed to address existing challenges through a privacy-conscious, scalable, and efficient solution while cultivating users' energy-saving habits through timely prompts to turn off unnecessary lighting. The proposed framework leverages Channel State Information (CSI) technology integrated with a multimodal ESP32-S3 IoT platform, designed to monitor indoor physical parameters including <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{CO}_{2}$</tex> levels, temperature, humidity, and illumination. The framework incorporates a comprehensive data processing pipeline featuring advanced signal processing techniques such as Hampel filtering, Kalman filtering, Wavelet denoising and Fast Fourier Transform analysis for CSI amplitude and phase data. Based on the processed time-frequency features, a DenseNet121-based deep learning architecture is constructed and compared with multiple baseline models. Evaluation results demonstrate that the proposed model achieves over 90 % classification accuracy in occupancy level prediction.