基于特征融合矩阵语音音色的厚薄度客观评价  

Objective Evaluation of Timbre Thickness for Speech Based on Feature Fusion Matrix

作　　者：涂中文[1] 张勤[2] 靳聪[2] 赵薇[2] TU Zhong-wen;ZHANG Qin;JIN Cong;ZHAO Wei(Broadcasting and Anchoring School,Communication University of China,Beijing 100024,China;Faculty of Science and Technology,Communication University of China,Beijing 100024,China)

机构地区：[1]中国传媒大学播音主持艺术学院,北京100024 [2]中国传媒大学理工学部,北京100024

出　　处：《北京理工大学学报》2018年第5期531-536,共6页Transactions of Beijing Institute of Technology

基　　金：国家自然科学基金重点资助项目(61631016);中国传媒大学交叉资助项目(3132017XNG1750)

摘　　要：提出一种基于特征融合矩阵的语音音色的厚薄度客观评价方法.提取最符合人耳听觉特性的梅尔频率倒谱系数(MFCC)和线性预测系数(LPC)作为语音特征参数,同时提取了3种非语音参数特征,包括共鸣因子指数、身高质量指数和肺活量体重指数,将这些特征进行融合即可形成特征融合矩阵,采用softmax分类器对语音音色中的厚薄度进行分级.实验结果表明,该方法可以获得较高的分级准确率.A new method was proposed based on a feature fusion matrix for objective evaluation of speech timbre thickness in this paper.The feature matrix was consisted of two parts,the speech feature parameter of MFCC and LPC,and the non-speech feature of resonance factor,body mass index,and vital capacity weight index.These features were fused in different combinations to establish a feature matrix.Finally,a softmax classifier was employed to classify the timbre thickness for speech into six levels.Experimental results show that the proposed method can achieve high performance for objective assessment.

关 键 词：特征融合矩阵 梅尔频率倒谱系数 线性预测 共鸣因子指数 厚薄度 

分 类 号：H01[语言文字—语言学]