中国科学院行为科学重点实验室知识库

摘要：许多研究发现音乐训练可以提高噪声中言语识别能力，为我们理解“鸡尾酒会问题”打开了一扇新的窗口。音乐训练可能通过两种途径产生这种优势效应，一种是自上而下地通过提高工作记忆、注意等高级认知或多通道信息整合能力来提升在噪音中分离和理解目标言语的能力；另一种是自下而上地通过增强对言语本身的表征能力由此提高噪声中言语识别的成绩。已有较多的研究就音乐训练自上而下的调节进行了探讨，但音乐训练对于言语声中不同成分的作用机制的研究还比较少，且主要研究手段为脑干频率追随反应——一种衡量听觉脑干对频率的精准加工能力的诱发电位。一段言语声波可以分离为慢速波动的时域包络（体现声波能量变化）和快速波动的时域精细结构的含的（包含声波频率成分和相位信息）。当前，普遍认为包络信息是决定言语识别的关键成分；而精细结构是一个相对冗余的信息，但在噪声掩蔽下的言语识别中起到了重要作用。本项研究探讨了音乐训练到底是提高了对包络信息本身的表征能力，还是通过提高对精细结构的加工能力来实现噪声中言语识别的优势效应。心理物理学实验中，匹配了工作记忆等认知能力的音乐家和普通被试在不同信噪比下分别复述去除包络或精细结构信息的语句。结果发现，在轻微噪声中（信噪比5dB）音乐家对保留精细结构（去除包络）的言语识别率要显著高于普通被试，而对保留包络（去除精细结构）的言语识别率没有任何差别。进一步，通过同时记录被试在聆听保留包络或精细结构言语声时的脑干频率跟随反应和皮层脑电并分别进行经验模态分析和互信息分析，结果发现，音乐家比普通被试在弱和无噪音条件下对保留精细结构的言语在脑干有更强的时间精细结构表征以及在皮层有更强的包络还原和同步能力构，并且这些指征与更好的噪音下言语识别成绩相关。这项研究提示，音乐家在噪声中可更好地利用言语时间精细结构这一线索以还原和同步言语的包络信息，从而提高噪音中的言语识别能力，这为音乐训练提高言语加工能力的认知神经机制提供了更加精细的解析。

Abstract：Many studies have found a musician advantage in identifying speech in noise (SIN), which opens a new avenue for studying the “cocktail-party problem”. Such an advantage might be achieved e by both top-down and bottom-up mechanisms: musical training could enhance higher-level cognitive functions (e.g., auditory attention or auditory working memory) and the cross-modal integration ability, or strengthen the processing of speech itself to improve speech in noise perception. While many research have focused on the higher-order functions in musicians, few studies have investigated how musical training influences the way that different acoustic components of speech been processed to improve the SIN performance. In those studies, the frequency-following response (FFR), an evoked response indexing how robust encoding of the periodicity of sound in the brainstem, had been used as an essential research method. A speech waveform could be decomposed into a slowly varying envelope (ENV) representing the fluctuation of energy and a rapidly varying temporal fine structure (TFS) containing the frequency and phase information. At present, it is generally believed that the envelope is most important for speech recognition, while the fine structure plays a key role in SIN although it is regarded as a redundant component in speech recognition in quiet environment. The current study investigated whether musical training enhances the processing of the envelope or fine structure of speech in adverse listening environment. In the psychophysical experiment，cognitively matched musicians and non-musicians were asked to repeat sentences that either the TFS or ENV has been removed and presented at different signal-noise-rate (SNR). Compared with non-musicians, musicians had significantly higher recognition of sentences preserving the TFS in mild noise (SNR = 5dB), and had no difference in repeating sentences without the TFS. Furthermore, electroencephalography from brainstem and cortex were recorded simultaneously when subjects listening to speech removed either the ENV or TFS, and analyzed by empirical mode decomposition and mutual information analysis, respectively. Results showed that, under quiet or weak noise conditions, compared to non-musicians, musicians were better at faithful representation of TFS in brainstem and reconstructing and synchronizing the speech envelope in auditory cortices from TFS-preserved speech, which was positively correlated with the better SIN performance. These findings suggest that, musical training may enhance the processing of the temporal fine structure of speech which allows strong recovery of, and neural synchronization with the speech envelope, which contributes to improved speech recognition in noisy environment. The current study provides a deeper understanding of the mechanisms on how musical training benefits speech processing.