Ear recognition based on compound structure classifier 基于主元分析與支持向量機的人臉識別方法
Based on the research of ear recognition with independent component analysis ( ica ) , a new compound structure classifier ( cscer ) ear recognition model was proposed 摘要在基于獨立分量分析的人耳識別方法研究基礎上,提出復合結構分類器的人耳識別通用模型。
This is the result of the reasons are : representative estimates of the number of structure classifiers the match capacity constraints and representative estimates of the number of structure is the indicator of grammar significance 造成這一情況的原因是:表約量的數(shù)量結構中量詞本身的搭配能力限制以及表約量的數(shù)量結構自身具有的特指的語法意義。
( 2 ) the influence to classification result is highly effected by using different classifier , for example , the center - vector algorithm obtains better classification results than other two algorithms . with the character feature , the average recall is 80 . 73 % , and the average precision is 82 . 94 % , and with the chinese - word feature , the average recall is 83 . 6 % , and the average precision is 85 . 97 % . different corpuses influence the classification result . for example , the average recall is 89 . 31 % and the average precision is 88 . 33 % , by using the news web pages as corpus from the web site " www . sina . com . cn " , which adopt the center - vector algorithm to structure classifier and select chinese - word as feature 對三種分類器分別以字、詞為特征進行分類測試、分析發(fā)現(xiàn):使用相同的分類算法,用詞作為特征項,比以字作為特征的分類效果好;用不同的算法構造分類器對分類效果的影響很大,如中心向量算法在字、詞特征下的分類效果優(yōu)于其他兩算法;在以字為特征的情況下,該算法的平均查全率80 . 73 ,平均查準率82 . 94 ;在以詞為特征的情況下,該算法的平均查全率83 . 6 ,平均查準率85 . 97 ;選用語料不同對分類效果也有影響,如用新浪網(wǎng)( www . sina . com . cn )網(wǎng)頁語料進行測試,使用中心向量法分類器和詞作為特征的情況下,平均準確率為89 . 31 ,平均查全率為88 . 33 。
My main work is as following : 1 ) applying feature mapping , sub - band structure classifier and multi - classifier cooperation to enhance the robust of system ; 2 ) giving out close - set fusion and open - set fusion functions to solve the problems of speaker identification and verification respectively ; 3 ) building the dynamic recognition length algorithm based on optimal stopping rules ; 4 ) developing a applied system based on the techniques above 主要工作是: 1 、提出參數(shù)映射、子帶結構分類器和多分類器系統(tǒng)以提高系統(tǒng)的魯棒性能; 2 、給出證據(jù)融合的閉集公式和開集公式,它們分別適用說話人辨認和確認問題的; 3 、通過最優(yōu)停止理論建立識別長度自適應算法; 4 、開發(fā)了一個實用的說話人識別系統(tǒng)。