基于波形相似叠加算法的多轨个性化耳鸣掩蔽康复音合成方法  

A sound synthesis method based on wareform similarity overlap-add for multi-track personalized tinnitus masking rehabilitation

作　　者：冯楚楠 何培宇[1] 徐瑞阳 方安成 潘帆[1] FENG Chunan;HE Peiyu;XU Ruiyang;FANG Ancheng;PAN Fan(College of Electronics and Information Engineering,Sichuan University,Chengdu 610065,China)

机构地区：[1]四川大学电子信息学院,成都610065

出　　处：《生物医学工程研究》2024年第5期385-393,共9页Journal Of Biomedical Engineering Research

基　　金：四川省自然科学基金项目(2022NSFSC0799)。

摘　　要：针对经典掩蔽疗法在耳鸣声治疗中难以兼顾康复音可听性与治疗有效性的问题,本研究基于波形相似叠加算法(waveform similarity overlap-add,WSOLA)和Loudnorm滤波器,提出了一种多轨个性化耳鸣掩蔽康复音(multi-track personalized tinnitus masking rehabilitation sound,MT-PTMRS)合成方法,以高效匹配个性化耳鸣频段,满足患者心理感知。本研究以生物自然音为核心,首先利用WSOLA进行音调变换,并通过反切迹强化个性化频段匹配效果;其次,基于满足患者主观偏好的环境自然音,使用迭代函数系统(iterated function system,IFS)得到相似不重复且时长可控的生物自然音;最后,使用Loudnorm滤波器统一其感知响度并采用渐变混合算法拼接。实验结果表明,本方法合成的康复音能有效匹配患者的耳鸣频段,符合1/f波动,康复音输入耳鸣模型能有效抑制对应频段耳鸣;试听结果表明,康复音悦耳流畅、舒缓自然。本研究方法对耳鸣声治疗研究具有一定的参考价值。Aiming at the problem that the classical masking therapy could not balance the of recovery sound audibility and the healing effectiveness in tinnitus treatment,based on waveform similarity overlap-add(WSOLA)and Loudnorm filter,we proposed a multi-track personalized tinnitus masking rehabilitation sound(MT-PTMRS)synthesis method to achieve efficient matching of personalized tinnitus frequency bands and satisfy patients'psychological perception.Based on natural sound,firstly,WSOLA was used to transform the tone,and the effect of personalized frequency band matching was enhanced by inverse notch.Secondly,based on the environmental natural sounds that met patients'subjective preferences,the iterated function system(IFS)was used to realize the similarity of biological natural sounds without repetition and the duration was controllable.Finally,the perceived loudness was unified by Loudnorm filter and splice by gradient mixing algorithm.The experimental results showed that the rehabilitation sound synthesized by this method could effectively match the tinnitus frequency band of patients and conform to the fluctuation of 1/f,and the input of rehabilitation sound into the tinnitus model could effectively suppress the tinnitus in the corresponding frequency band.The results showed that the sound of rehabilitation was pleasant and smooth,soothing and natural.The proposed method has certain reference value for the research of tinnitus sound treatment.

关 键 词：耳鸣声疗法 音频变调 掩蔽治疗 自然声 频率匹配 声音合成 

分 类 号：R318[医药卫生—生物医学工程] TN912[医药卫生—基础医学]