基于数理方法的语音重构研究  

A Study on Speech Reconstruction Based on Mathematical Methods

作　　者：王咿卜 李建文[2] 王术 WANG Y-bo;LI Jian-wen;WANG Shu(College of Mathematics and Computer Application/Engineering Research Center of Qinling Health Welfare Big Data,Universities of Shaanxi Province,Shangluo University,Shangluo 726000,Shaanxi;School of Electronic Information&Artificial Intelligence,Shaanxi University of Science&Technology,Xi′an 710021,Shaanxi)

机构地区：[1]商洛学院数学与计算机应用学院/秦岭康养大数据陕西省高校工程研究中心,陕西商洛726000 [2]陕西科技大学电子信息与人工智能学院,陕西西安710021

出　　处：《商洛学院学报》2023年第2期62-70,共9页Journal of Shangluo University

基　　金：国家自然科学基金重点项目(60672001);商洛学院科研基金项目(22SKY001);商洛学院重点学科建设项目。

摘　　要：为了维护听障患者的身心健康,探究弱听患者在不同频率段存在的分贝值衰减的现象,从语音的发音原理出发,采用自相关函数法、倒谱法分别对语音信号进行基频、共振峰提取。通过语谱图探究基频曲线与共振峰的相关性,采用曲线函数拟合法对基频进行拟合,采用帧重叠相加法实现语音合成。进一步调整特征参数值实现语音重构,总结听障患者在不同频段的分贝值变化规律。结果表明,对于语音重构函数,在特定频率段内改变分贝值能够有针对性地提升听障患者的听力效果。In order to maintain the physical and mental health of hearing impaired patients,and explore the phenomenon of attenuation of decibel values in different frequency bands for patients with hearing impairment,starting from the pronunciation principle of speech,the autocorrelation function is used to extract the fundamental frequency of the speech signal,and the cepstrum is used to extract the formant of the speech signal.The correlation between the fundamental frequency curve and the formant is explored through the spectrogram.The fundamental frequency is fitted by the curve function fitting method,and the speech synthesis is realized by the frame overlap addition method.Further adjust the value of the feature parameters to realize speech reconstruction,and summarize the change rule of decibel value of hearing impaired patients in different frequency bands.

关 键 词：听障患者 基频曲线 共振峰 帧重叠相加法 语音重构 

分 类 号：TN912.33[电子电信—通信与信息系统]