Communication-aided Terahertz Sensing: A Novel Indoor People Counting System Via Deep Learning
Eslam Hasan, Elmahedi Mahalal, Muhammad Ismail, Zi-Yang Wu, Mostafa M. Fouda, Zubair Md. Fadlullah
Abstract
Indoor people counting systems are used in security monitoring, energy management, room resources adjustment, market research, and smart homes. However, the existing radio-frequency-based indoor people counting systems use a different frequency than the utilized radio frequency communication signal, which adds more costs for system deployment and wastes the radio frequency resources. This paper introduces a novel communication-aided terahertz (THz) sensing approach for robust indoor people counting that utilizes the THz communication downlink signal to sense the number of indoor people. Leveraging the wireless THz communication downlink signal, we propose a device-free, cost-effective, and non-intrusive indoor people counting system. Our method employs a 1D convolutional neural network (CNN) to process historical THz downlink channel gain data and accurately estimate the number of indoor occupants. The numerical results demonstrate the effectiveness of the proposed approach, achieving a remarkable 99.5% accuracy in people counting indoors up to eight people. The proposed model's ability to maintain high accuracy in indoor people counting across different numbers of users demonstrates its effectiveness and robustness in capturing the occupancy signature from the wireless THz downlink communication signal in indoor environments. Also, the accuracy of the proposed CNN time series classifier outperforms the random forest times series classifier with the catch22 feature extractor by more than 10% without needing any feature extraction methods. To the best of the authors' knowledge, this study represents the first investigation into indoor people counting in the THz frequency band utilizing the THz downlink communication signal for sensing the number of indoor occupants.