透過您的圖書館登入
IP:44.223.39.199
  • 學位論文

基於光體積描述訊號之長期監控心智負荷偵測系統

PPG-based Long-term Mental Workload Detection System

指導教授 : 吳安宇
若您是本文的作者,可授權文章由華藝線上圖書館中協助推廣。

摘要


隨著人類社會不斷進步,我們的生活品質獲得了提升,但同時心理健康問題如 憂鬱症、焦慮症等心理疾病相繼產生,並成為全世界需要面對的問題,而心理健康 相關問題的一大來源為工作壓力。心智負荷 (MentalWorkload) 被認為是個體可用 的資源量與作業情況所要求資源量之間的差異,較高的心智負荷意味在處理工作 時需消耗更大量的訊息處理能力,因此透過生理訊號評估心智負荷可即時反映工 作時的狀態,避免工作量超出自身負荷,並藉由生理訊號的長期監控維持心理健康。 然而,需達成長時間的監控必須透過穿戴式裝置,由於穿戴式裝置的便利性使 得活動不會受到限制,造成收錄到的訊號產生不穩定的情形,進而降低分類表現。 因此,我們在此論文提出了一個穩健的心理負荷偵測系統,透過基線飄移的去除及 非線性的特徵抽取提高對噪聲的容忍度。接著基於原本的系統,開發了針對光體積 描述訊號的輔助前處理系統,藉由心電訊號的標記及機器學習的輔助,能夠準確的 識別穿戴式中常見的運動偽影並進行去除,大幅降低穿戴式裝置不穩定所產生的 誤差。接著,將針對此誤差進行評估,在特徵值產生誤差的情況下,勢必會使分類 器產生不確定性,本文針對此不確定性提出了一個評估系統,將此不確定性量化為 可能分類錯誤的機率,藉此找出一些高機率分類錯誤的資料並將其去除,進而減少 誤判的情形。在此架構下,用於穿戴式裝置的光體積描述訊號的分類準確度可以大 幅提升,並能逼近心電訊號的分類表現。

並列摘要


Along with the economic development and social progress, the quality of our life has been improved. However, the mental health problems such as depression, anxiety, and other mental disorders have emerged and become a global problem. A major source of mental health problems comes from the stress of work. Mental workload is considered to be the difference between the amount of resources available to an individual and the amount of resources required by operating a task. A high mental workload means a more significant amount of information processing capacity or resources in performing a task. Therefore, the assessment of mental workload can reflect the state of work in real-time through physiological signals to avoid overload and maintain mental health through continuous monitoring. However, long-term monitoring must be achieved with a wearable device. The use of wearable devices is not limited by activity, resulting in the stability problem that reduces the classification performance. Therefore, we propose a robust mental workload detection system that improves noise tolerance through baseline removal and nonlinear feature extraction. In addition, based on the system, we develop an auxiliary preprocessing system, which can accurately identify and remove motion artifacts in wearable devices with the assistance of ECG labels and machine learning. It significantly reduces the error caused by motion. After the artifact removal, we proposed an evaluation method that quantified the uncertainty as the probability of misclassification due to the loss of information caused by artifact removal. The accuracy of the proposed system can be significantly improved and can approximate the classification performance of the ECG signal.

參考文獻


[1] Roosa Tikkanen et al., “Mental Health Conditions and Substance Use: Comparing U.S. Needs and Treatment Capacity with Those in Other High-Income Countries.” (Commonwealth Fund, May 2020). https://doi.org/10.26099/09ht-rj07
[2] Harnois, Gaston, and Phyllis Gabriel. “Mental health and work: Impact, issues and good practices.” (2000).
[3] Brundtland, Gro Harlem. “Mental health in the 21st century.” Bulletin of the world Health Organization 78 (2000): 411-411.
[4] National Alliance on Mental Illness. (2017 Jan.26). “How The Affordable Care Act Helps Americans With Mental Illness And Why It Matters.” https://www.nami.org/About- NAMI/NAMI-News/2017/What-s-At-Stake
[5] Gopher, D., and Donchin, E. (1986). “Workload: an examination of the concept,” in Handbook of Perception and Human Performance, Vol. 2: Cognitive Processes and Performance, eds K. R. Boff, L. Kaufman, and J. P. Thomas (Oxford: John Wiley Sons), 1–49.

延伸閱讀