融合AAM、CNN與LBP特征的人臉表情識別方法
發(fā)布時間:2018-08-01 11:05
【摘要】:提出一種人臉表情識別方法,融合主動形狀模型(AAM)、卷積神經(jīng)網(wǎng)絡(luò)(CNN)和局部二元模式(LBP)3種特征區(qū)分不同表情。進行圖像預(yù)處理操作,核心是采用AAM方法進行姿態(tài)校正與面部裁剪,得到規(guī)范化的表情圖像,在全圖上提取AAM和CNN兩組全局特征,在AAM定位的6個人臉局部區(qū)域圖像上提取LBP局部特征,融合全局特征和局部特征,采用隨機森林方法進行特征分類。在Cohn-Kanade數(shù)據(jù)集上的實驗結(jié)果表明,該方法的表情識別率高,是一種有效的表情識別方法。
[Abstract]:A facial expression recognition method is proposed, which combines the active shape model (AAM),) convolution neural network (CNN) and the local binary pattern (LBP) to distinguish different facial expressions. The core of image preprocessing is to use AAM method for attitude correction and facial clipping, to get standardized facial expression image, and to extract two groups of global features of AAM and CNN on the whole image. The local features of LBP are extracted from the local images of 6 faces located by AAM, and the global features and local features are fused, and the feature classification is carried out by using the stochastic forest method. The experimental results on the Cohn-Kanade dataset show that this method has a high expression recognition rate and is an effective expression recognition method.
【作者單位】: 河南工程學(xué)院計算機學(xué)院;鄭州大學(xué)軟件與應(yīng)用技術(shù)學(xué)院;
【基金】:國家社會科學(xué)基金項目(15XTQ010)
【分類號】:TP391.41
[Abstract]:A facial expression recognition method is proposed, which combines the active shape model (AAM),) convolution neural network (CNN) and the local binary pattern (LBP) to distinguish different facial expressions. The core of image preprocessing is to use AAM method for attitude correction and facial clipping, to get standardized facial expression image, and to extract two groups of global features of AAM and CNN on the whole image. The local features of LBP are extracted from the local images of 6 faces located by AAM, and the global features and local features are fused, and the feature classification is carried out by using the stochastic forest method. The experimental results on the Cohn-Kanade dataset show that this method has a high expression recognition rate and is an effective expression recognition method.
【作者單位】: 河南工程學(xué)院計算機學(xué)院;鄭州大學(xué)軟件與應(yīng)用技術(shù)學(xué)院;
【基金】:國家社會科學(xué)基金項目(15XTQ010)
【分類號】:TP391.41
【相似文獻】
相關(guān)期刊論文 前10條
1 王志良,陳鋒軍,薛為民;人臉表情識別方法綜述[J];計算機應(yīng)用與軟件;2003年12期
2 孫蔚;王波;;人臉表情識別綜述[J];電腦知識與技術(shù);2012年01期
3 楊梅娟;;人臉表情識別綜述[J];甘肅科技;2006年04期
4 劉曉e,
本文編號:2157368
本文鏈接:http://sikaile.net/kejilunwen/ruanjiangongchenglunwen/2157368.html
最近更新
教材專著