时域精细结构与包络信息在噪声下声调识别中的作用  

The Contributions of Temporal Fine Structure and Envelope Information for Lexical Tone Identification in Noise

作　　者：亓贝尔 刘佳星 古鑫 刘博 Qi Beier;Liu Jiaxing;Gu Xin;Liu Bo(Beijing Tongren Hospital,Capital Medical University,Beijing Institute of Otolaryngology,Key Laboratory of Otolaryngology Head and Neck Surgery(Capital Medical University),Ministry of Education,Beijing,100730,China;Beijing Chaoyang Hospital,Capital Medical University,Beijing,100730,China)

机构地区：[1]首都医科大学附属北京同仁医院,北京市耳鼻咽喉科研究所,耳鼻咽喉头颈科学教育部重点实验室(首都医科大学),北京100005 [2]首都医科大学附属北京朝阳医院

出　　处：《听力学及言语疾病杂志》2022年第1期1-4,共4页Journal of Audiology and Speech Pathology

基　　金：北京市卫生系统高层次卫生技术人才培养计划(2015-3-019)。

摘　　要：目的探讨时域精细结构信息(temporal fine structure,TFS)和时域包络信息(envelope,Env)在噪声下声调识别中的作用。方法自行编制噪声下声调识别能力测试材料,对20例年龄19~30岁、母语为汉语普通话的听力正常人进行五种信噪比(SNR)条件下(SNR分别为-18、-12、-6、0、+6 dB)的声调识别能力测试,采用广义线性模型(generalized linear model,GLM)对所得数据进行统计分析。结果①噪声环境下的声调识别与TFS信息和Env信息均相关,两者协同作用更有助于提高噪声下的声调识别能力,在言语谱噪声(speech spectrum-shaped noise,SSN)条件下Env、TFS以及二者协同作用与声调识别成绩的回归系数分别为0.095(P<0.0001)、0.070(P<0.0001)和-0.002(P<0.0001);在两人谈话噪声(two-talker babble,TTB)条件下Env、TFS以及二者协同作用与声调识别成绩回归系数分别为0.052(P<0.0001)、0.073(P<0.0001)和-0.0003(P=0.13)。②当TFS信息和Env信息量相等时,改善信噪比有利于提高声调识别成绩;SNR_(TFS)和SNR_(Env)相等时,SSN噪声下五种信噪比时声调识别平均正确率分别为27.6%、60.2%、82.1%、93.9%和94.7%;TTB噪声下五种信噪比时声调识别平均正确率分别为53.5%、72.0%、86.4%、92.7%和95.0%。结论时域精细结构信息和时域包络信息对于听力正常人进行噪声下声调识别具有同等作用,两者协同作用更有助于提高噪声下的声调识别能力。Objective To investigate the role of temporal fine structure(TFS)and envelope(Env)cues in tone perception in noise.Methods Twenty normal hearing listeners aged ranged from 19 to 30 yr(mean age=23.71 yr,SD=2.995)participated in this study(10 females,10 males).All participants were native speakers of Mandarin.The tone perception test used a word-based,four-alternative forced-choice test(4-AFC)paradigm controlled using custom MATLAB routines running on PCs.Targets and the masker were added at five different signal-to-noise ratios(SNRs=-18,-12,-6,0,+6 dB).Generalized Linear Model(GLM)was used to examine the effects of TFS and Env on tone recognition performance.Results①The tone-recognition performance improved as the SNRs in either TFS or E increased.Both TFS and E contributed significantly to tone recognition in various SNR.In the SSN conditions,the coefficients(β)of the GLM fitting for SNR in TFS,SNR in Env,and the interaction between SNRs in TFS and Env were 0.095(P<0.0001),0.070(P<0.0001)and-0.002(P<0.0001),respectively.In the TTB conditions,the coefficients(β)of the GLM fitting for SNR in TFS,SNR in Env,and the interaction between SNRs in TFS and Env were 0.052(P<0.0001),0.073(P<0.0001)and-0.0003(P=0.13),respectively.②When the SNRs in both TFS and Env were equal,the tonal perception results were increased withimproving SNR.The average tone-recognition scores were 27.6,60.2,82.1,93.9,and 94.7%correct in SSN conditions at SNR of-18,-12,-6,0,and+6 dB,respectively and were 53.5,72.0,86.4,92.7,and 95.0%correct in TTB conditions at the same SNR.Conclusion The present data have showed that both TFS and Env cues contributed to lexical tone perception in noise.Therefore,providing both TFS and Env in hearing devices such as cochlear implants should be beneficial for the hearing-impaired listeners in everyday noisy environment.

关 键 词：声调 时域精细结构 时域包络 

分 类 号：R339.16[医药卫生—人体生理学]