利用前額訊號偵測駕駛者在虛擬駕車環境中的清醒程度

Abstract

根據過去研究顯示，跟人類打瞌睡有很強的關聯性的腦區主要是跟枕葉區的腦波變化(8~12Hz) 有很高的相關。然而，若要在真實生活中要使用傳統的電極帽來擷取腦電波訊號是相當不方便的，而且需要繁複的準備流程，對駕駛者也相當困擾。因此本論文的主要目的就是要使用前額的電極來偵測駕駛者的精神狀態並且預測開車軌跡，這樣一來我們就能在真實的生活當中建立一個較為可行的偵測系統。

本論文主要分析並且比較了分別來自於枕葉區還有額葉區的腦波訊號。而來自於這些區域的頻率能量變化將會被用來當作輸入參數並且使用線性回歸模型來預測駕駛者的開車軌跡。而根據我們的結果，使用來自枕葉區的訊號可以得到最高的預測率。

此外，我們也證明了有一個跟打瞌睡相關的腦區位於額葉區，結果顯示4∼7Hz 頻帶的變化跟打瞌睡有很高的相關性。而使用前額電極的訊號來預測開車軌跡，雖然比使用枕葉區的準確率來的差一些但仍有八成的準確率，但這意味著使用前額電極的訊號就可以預測駕駛者的精神狀態，如此一來就可以省下使用電極帽而帶來的複雜準備程序，這樣的偵測系統在未來就可以被廣泛且容易的被使用在真實的駕車環境當中。

Previous studies showed that the alpha power increases in the occipital lobe highly related to human drowsiness. However, the acquisition of occipital EEG signals with the traditional electrode cap is inconvenient. Thus, the main purpose of this study was to confirm whether the forehead EEG signals could reflect the driver’s drowsiness and be able to use to estimate driver’s driving trajectory for constructing a feasible detecting system that can be applied in real life. Brain signals acquired from the occipital and the frontal lobe were analyzed and compared in this study. The frequency power changes in these components were used as features and fed into linear regression model to predict driver’s driving performance. Results showed the highest estimation accuracy was yielded with the features extracted from the occipital ICs cluster.

We also found that there is another drowsiness-related brain source located in the frontal lobe. Furthermore, the increases of the theta power in the frontal lobe also highly correlated to the driver’s drowsiness. Comparing the conventional methods using the occipital activities, the estimation accuracy using the forehead signals is slightly lower but the estimation accuracy was still higher than 0.8.

Results demonstrated that forehead signals could be used to estimate the drivers’ drowsiness. The new detecting system, using forehead signals, not only can correctly estimate the user’s drowsiness but also can drastically reduce the preparation time. In the future, such detection system will be easily and widely applied in the real operational environments.