首页>
外国专利>
THE METHOD TO IMPROVE THE PERFORMANCE OF SPEECH/MUSIC CLASSIFICATION FOR 3GPP2 CODEC BY EMPLOYING SVM BASED ON DISCRIMINATIVE WEIGHT TRAINING
THE METHOD TO IMPROVE THE PERFORMANCE OF SPEECH/MUSIC CLASSIFICATION FOR 3GPP2 CODEC BY EMPLOYING SVM BASED ON DISCRIMINATIVE WEIGHT TRAINING
展开▼
机译:基于区分权重训练的SVM应用SVM提高3GPP2编解码器语音/音乐分类性能的方法
展开▼
页面导航
摘要
著录项
相似文献
摘要
PURPOSE: a kind of background model 3GPP2 codecs improving classification performance are not provided to the feature vector of different coating weight values by a SVM (support vector machine) based on differentiation weight training, it is input to the influence degree of SVM, according to the classification voice/music of feature vector. ;CONSTITUTION: the SVM codecs of the coded portion in preprocessing process, at least one feature vector are extracted (S100). A kind of voice/music Classification and Identification formula is drawn by using lagrangian optimization method and extracted feature vector (S200). The weight is obtained to the characteristic value of the extraction, it is contemplated that GPD (General Probability decline) base MCE (minimum classification mistake) training (S300). It is applied to the voice/music Classification and Identification formula drawn to resulting weighted value. The voice signal is input to svm classifier (S400). ;The 2011 of copyright KIPO submissions
展开▼