智慧手機結合G-sensor之打瞌睡偵測系統之研發

Abstract

由於紅外線車上型打瞌睡偵測器有照久了眼睛會有灼熱感的情形，故本論文的方向為開發不用紅外線且在光線不足時能做偵測打瞌睡的系統，我們用G-sensor偵測瞌睡點頭來補足光線不足時的情況。 為了解決問題，本研究開發出打瞌睡偵測器App和G-sensor帽子，App結合智慧手機內建的攝影機用來偵測瞌睡的閉眼，App結合G-sensor帽子則可偵測瞌睡的點頭，閉眼和點頭的偵測是同時進行，所以當光線不足，無法偵測閉眼時，還有點頭偵測判斷使用者是否在打瞌睡，App偵測閉眼部分的演算法用的是Haar cascade演算法，平均處理1張影片要0.48秒，App偵測點頭部分，G-sensor帽子上裝有G-sensor、Arduino板子、藍芽模組及行動電源，G-sensor的類比資料會先透過Arduino板子轉成字串數據，再透過藍芽傳送到App做分析。 由我們的實驗一得知，偵測閉眼的準確率為99.52%，偵測點頭的準確率為100%，由我們的實驗二得知，偵測閉眼的準確率為99.89%，偵測點頭的準確率為100%，由於兩實驗偵測點頭的準確率都為100%，故能解決光線不足時的問題。

關鍵字：打瞌睡偵測器、G-sensor、Haar cascade演算法、瞌睡的點頭、瞌睡的閉眼

This research has developed a system to detect doze (sleep) nod while driving, through G-sensor and without infrared. This newly invented system is not only able to detect slim eyes-brainwave-and head motion of the drivers at dim light, but capable to eliminate allergy to the eyes that are caused by infrared eye detectors, as the primary source of light for most detection equipment used by many of the eye tracking systems currently available in the market. To solve the problem, the research develop doze detector by using App and G-sensor cap, App combine smart phone built-in camera to detect sleepy eyes closed, App can be combined with G-sensor cap detects sleepy nod, eyes closed and nodded detection are performed simultaneously, so when lighting is poor, can not detect when eyes closed, and determine whether the user is in a nod to detect doze. To detect doze (sleep) eye closure by APP, this research has adopted Haar Cascade algorithm for measurement, with average processing time of 0.48 seconds per one photo. To effectively assess doze (sleep) nod, a G-sensor hat has been developed to cooperate with APP detector. The G-sensor cap comes with G-sensor device, Arduino board, Bluetooth module, and portable battery. G-sensor analog data will be converted into a string of data through the Arduino board, then sent via Bluetooth to the App for analysis. By our first experiment, detection accuracy rate of 99.52 percent with eyes closed, nodding detection accuracy was 100%. By our second experiment, detection accuracy rate of 99.89 percent with eyesclosed, nodding to detect accuracy was 100%, due to the accuracy of the two experimental detection nod are 100%, it can solve the problem when lighting is poor.

Keyword: Doze (sleep) nod detector, G-sensor, Haar cascade algorithm, Doze (sleep) nod, Doze (sleep) eye closure