運用智慧型手機及Wi-Fi 訊號之非協同式排隊辨識

Abstract

排隊行為是人類生活中常見的一種群體行為，在許多場合中人們經常花費大量的時間在排隊等待上，例如：電影院購票，遊樂園入場，速食店等候購餐等。排隊偵測是近年來一門新興的研究議題，先前研究的方法主要目是利用感測技術或者影像處理的方式偵測人類排隊的行為，但容易因使用程式者的比例不足或者影像中人物重疊，導致辨識準確率不佳。本研究提出利用手機上的三軸加速度計以及Wi-Fi訊號，分析排隊時人們特有的行為特徵，並運用資料探勘技術分析感測資料以偵測使用者是否正在排隊，再將排隊者利用不同隊伍之間等候時間的差異進行分群，最後透過Wi-Fi訊號強度分析隊伍長度以及等候時間。本論文提出的架構能夠有效率的辨識使用者排隊情況並且提供排隊者預估等候時間。為了評估我們的系統，真實生活的排隊場境中進行辨識排隊實驗，也將整個系統實作在安卓智慧型手機上來驗證我們的方法在目前裝置上運行的可行性，實驗結果顯示，準確率優於其他排隊辨識系統。

Queuing behavior is an important kind of group activities in human’s life. In many real-life scenarios like movie theaters, amusement parks, and restaurant, people spend a substantial amount of time waiting on lines. Detecting the status of the queuing behavior may benefit both customers and businesses. In view of this, a new research topic called queuing detection has emerged in recent by machines. Providing such queuing information helps customers better spend time doing something alternative rather than blindly waiting in line, thereby mitigating their anxiety and improving their experience. In addition, managing queues is important for business since it can help reduce inefficient resource allocation and revenue loss. Therefore, there is a need for a better understanding of emerging trends of the queues in order to not only improve user experience but also benefit business. We implement a prototype on Android platforms using widely available sensors such as accelerometer and gyroscope on mobile phone. Our method not only has achieved high precision and estimate the waiting time efficiently.