- 學位論文
基於Wi-Fi通道狀態資訊與卷積神經網路之室內人員偵測
Indoor People Detection Based on Channel State Information and Convolutional Neural Network
摘要
本論文使用ESP32配合Android智慧型手機收發Wi-Fi訊號，採集通道狀態資訊(Channel State Information, CSI)，分析在相同環境、不同狀態下CSI產生的變化。本研究提出了一個基於CNN的室內人員偵測系統，該系統可以實現室內人員的偵測區分出室內是否有人員存在。其基本思路是將CSI資料作為輸入，通過CNN模型學習到特徵，從而對室內人員進行分類。實驗結果顯示，本系統具有很好的性能，能夠有效地對室內人員的存在進行偵測。另外，Wi-Fi感測還可以應用於運動偵測、睡眠監測等領域，這些應用場景將會在未來得到更廣泛的應用。Wi-Fi感測技術仍然存在著一些問題和挑戰，例如，Wi-Fi訊號受到環境干擾和訊號遮蔽等因素的影響，本論文也針對這個部分設計實驗，分析與討論辨識結果。整體上來看，Wi-Fi感測技術應用於室內人員偵測是可行並且有極佳的辨識正確率。
並列摘要
This thesis proposes a CNN-based indoor personnel detection system that uses an ESP32 in conjunction with an Android smartphone to send and receive Wi-Fi signals, collecting Channel State Information (CSI) to analyze the variations in CSI under different states in the same environment. The system can distinguish whether there are individuals present indoors . The fundamental approach is to use CSI data as input, allowing the CNN model to learn features and classify indoor personnel. Experimental results show that the system performs well, effectively detect the presence of indoor individuals. Furthermore, Wi-Fi sensing can also be applied in various fields such as motion detection and sleep monitoring, and these applications will be more widely used in the future. However, Wi-Fi sensing technology still faces some challenges, such as the influence of environmental interference and signal obstruction. The thesis addresses this issue by designing experiments to analyze and discuss recognition results. Overall, the application of Wi-Fi sensing technology in indoor human detection proves to be feasible with excellent recognition accuracy.