基于自适应变分模态分解的佤语孤立词共振峰估计  

Formant estimation of isolated words in the Wa language based on adaptive variational mode decomposition

作　　者：杨建香 佘玉梅[1] 傅美君 和丽华 解雪琴 潘文林[1] YANG Jian-xiang;SHE Yu-mei;FU Mei-jun;HE Li-hua;XIE Xue-qin;PAN Wen-lin(School of Mathematics and Computer Science, Yunnan Minzu University, Kunming 650500, China;School of Information Science and Technology, Southwest Jiaotong University, Chengdu 611756, China)

机构地区：[1]云南民族大学数学与计算机科学学院,云南昆明650500 [2]西南交通大学信息科学与技术学院,四川成都611756

出　　处：《云南民族大学学报（自然科学版）》2019年第3期297-305,共9页Journal of Yunnan Minzu University:Natural Sciences Edition

基　　金：云南民族大学研究生创新基金(2018YJCXS223)

摘　　要：基于变分模态分解算法实现对语音共振峰的提取,针对其存在共振峰合并和虚假峰值2个主要问题,通过对分解模态数、平衡约束参数的分析,提出了自适应变分模态分解法.并从的正交性、能量保存度2个方面证明了该方法的可行性;最后,利用该方法实现对佤语共振峰的估计.实验结果表明,基于自适应变分模态分解对佤语孤立词的共振峰估计平均正确率可达85.50%.As for the two major problems of formant merger and false peaks, the algorithm based on variational mode decomposition helps realize the extraction of the formants from speech signals. This paper proposes a method of adaptive variational mode decomposition (AVMD) by analyzing number K of the decomposition mode and parameter(α)as the equilibrium constraint. The feasibility of the method is respectively analyzed from two aspects of the IMF: its orthogonal performance, and its energy conservation. Finally, it realizes the estimation of formant parameters of the Wa speech. Experimental results show that based on the adaptive variational mode decomposition for estimating the formant of the isolated words of the Wa language, the average accuracy rate can reach 85.50%.

关 键 词：变分模态分解法(VMD) 本征模态函数(IMF) 分解模态数(K) 平衡约束参数(α) 自适应变分模态分解法(AVMD) 

分 类 号：TN912.34[电子电信—通信与信息系统]