中原大學電機工程學系
Loading
專題成果首頁>專題成果
107學年度 >訊號與系統組
智慧影像人臉情緒辨識
指導老師:王佳盈   組長:歐陽百祐   組員:邱耀霆、歐晉豪
某些過動兒或自閉症童在判斷別人的情緒上,比起一般正常顯得較為困難,他們無法從別人臉部的表情了解對方目前所要達緒。本專題主要的目,就是希望幫助或訓練這些特殊兒童透過人臉辨識技術,感測對方的喜怒哀樂然後用其他式把訊息回饋給這些兒童。
判斷臉部情緒有許多方法,在人偵測上我們使用OpenCV中內建中內建的Haar級聯分類器 (Haar cascade)進行偵測,而人臉情緒識別則是使用深度學習的開源軟體庫Tensorflow,並使用網路上的情緒資料庫以及自己蒐網路上的情緒資料庫以及自己蒐集的樣本,然後將這些樣本丟進CNN(Convolution Neural Network)神經網路模型進行訓練,以產生預測並藉此來幫助我們辨別情緒。
有了可以判斷人臉情緒的工具後,我們將這些預測模型放入樹莓派所做的小型裝置中,以方便隨身使用。此是個模擬在未來發展上希望能夠做出類似google glass的介面並且做到真正隨身攜帶。
Hyperactive children or autistic children are more difficult to judge other people's emotions than normal people. They can't understand the emotions of the other party by observing other people's faces. The main purpose of this project is to help or train these special children. We hope that through the technique of face emotion recognition, the other person's emotions can be detected, and then the message can be sent to these children in other ways.
There are many ways to judge facial emotions. Basically, we use the Haar cascade built in OpenCV to detect human face. Then we use Tensorflow to tarin a model for detecting face emotions. We use the online face emotional database together with and the samples collected by ourselves. By feeding these samples into the CNN (Convolution Neural Network) neural network model, we get a predictive model which can be used to identify face emotions. We then put the predictive model into a small device built by using a Raspberry Pi. This device is used as an alternative for experiment. We hope that in the future, the model can be put into a device like the google glass such that it is portable
我們決定將壓電陶瓷晶片和鞋子做結合。
   
 
32023 桃園市中壢區中北路200號 No-.200, Zhongbei Rd., Zhongli City, Taoyuan County 320, Taiwan(ROC) TEL:03-2654801 FAX:03-2654899