心跳感測器輔助影像深度學習應用於臉部痛苦指數之判別

Heartbeat sensor-assisted image depth learning applied to the identification of facial pain index

陳梓瑄(Chen Zi-Xuan)

指導教授 ： 陳美勇

國立臺灣師範大學/科技與工程學院/機電工程學系/碩士(2018年)

https://doi.org/10.6345/THE.NTNU.DME.016.2018.E08

摘要

本研究提出了一種估計人臉疼痛強度的方法。使用回歸卷積神經網路訓練模型，其中包含三層卷積層及三層池化層。此外，使用心跳感測器幫助臉部疼痛識別，目的是更準確地判斷人的疼痛程度。在兩個感測器的偵測及相互輔助下，可以極大程度的預防危險性的發生。例如使用在跑步、復健及醫療上，若能夠第一時間偵測到使用者的痛苦指數及心跳數異常，將能有效且迅速的做第一時間的處理。同時，本論文另一個貢獻為根據醫療及運動等相關文獻做出實際測試結果，將心跳感測與臉部疼痛做出結合與應用，並且能實際應用於生活當中。
本研究結果顯示，i5雙核計算機的MSE為0.11，Pearson相關係數接近1（r = 0.98），平均運算速度達到70 FPS。除了能夠高速運算臉部痛苦指數，也能迅速對硬體下達指令。

關鍵字

卷積神經網路 ； 電腦視覺 ； 心跳感測器

並列摘要

In this paper, a method for estimating the intensity of pain in human face is presented. Human face image is extracted by using Convolution Neural Network (CNN) and max pooling. Both facial image and pain index will be computed by the regression Convolution neural network, whose training results are verified. In addition, used a heartbeat sensor to assist facial pain recognition in order to more accurately detect pain in a person. The detection of two sensors can greatly prevent the occurrence of danger. For example, in running, rehabilitation and medical care. If we can detect the user's condition for the first time, we will be able to deal with it effectively and quickly. At the same time, another contribution of this paper is to make the actual test results according to the medical and sports related paper. It combined the heartbeat sensation with facial pain, and can be applied in real life.
In terms of accuracy, the mean squared error is 0.11, and the correlation coefficient is 0.98. In terms of execution speed, the average computing speed achieves 75 FPS on the i5 dual-core computer. In addition to being able to calculate face pain index at high speed, it can also quickly send instructions to hardware.

並列關鍵字

Convolution Neural Network ； Computer vision ； Heartbeat sensor

參考文獻

一、中文文獻

『耐力網心率區間檢測』, available: http://center4gaming.org/c4g/index.php/estimate/index . July.2018.

Lin Kao-Yuan, “Implemented Rapid Pain Intensity Estimation from Facial Image using Artificial Neural Network,” 台灣台北: 國立臺灣師範大學機電工程學系碩士論文,2016.

二、英文文獻

"Facial Action Coding System," Wikimedia Foundation, Inc, [Online]. Available: https://en.wikipedia.org/wiki/Facial_Action_Coding_System. [Accessed 25 2 2016].