共振峰信息在汉语声调感知中的作用  

Role of formants in Mandarin lexical tone perception

作　　者：王硕[1] Robert Mannell Philip Newall 董瑞娟[1] 李靖[1] 张华[1] 陈雪清[1] 韩德民[1] 

机构地区：[1]北京市耳鼻咽喉科研究所首都医科大学附属北京同仁医院耳鼻咽喉头颈外科耳鼻咽喉头颈科学教育部重点实验室(首都医科大学),北京100730 [2]澳大利亚麦考里大学语言学系 [3]澳大利亚听力研究合作中心 [4]澳大利亚皇家盲聋儿童研究所

出　　处：《中国耳鼻咽喉头颈外科》2012年第1期8-11,共4页Chinese Archives of Otolaryngology-Head and Neck Surgery

基　　金：国家自然科学基金项目(81070784);北京市卫生系统高层次卫生技术人才培养基金项目(2009-3-29);北京市自然科学基金(7122034);首都医科大学附属北京同仁医院科研基金资助课题(2012-YJJ-003)联合资助

摘　　要：目的探讨言语频域共振峰信息在汉语普通话声调识别中的作用。方法采用线性预测编码(linearprediction coding,LPC)技术将单音节字的频域包络(即共振峰信息)与精细结构(即基频谐波信息)分离,合成128个只具有频域包络信息的"单音节字"。使用这些"单音节字"测试20例听力正常受试者的声调感知能力。结果听力正常受试者依靠共振峰信息识别三声的正确率为69%,识别二声的正确率为42%,识别一声和四声的能力均低于机会值(25%),声调的平均识别率为36%。只依靠共振峰信息,三声与二声的识别容易混淆,一声与四声较易误识别成二声与三声。结论共振峰信息(尤其是第一共振峰)可以为声调识别提供一些线索,但不起主要作用。OBJECTIVE This study aimed to evaluate the role of formants in Mandarin lexical tone perception.METHODS The Linear Predictive Coding technique was applied to synthesize 128 monosyllabic tokens only having formant information.Twenty normalhearing listeners participated in this study.They were measured tone perception performance using 128 synthetic tokens.RESULTS The correct percentage of tone perception to Tone 3 and Tone 2 was 69% and 42% respectively,while the percentage of tone perception to Tone 1 and Tone 4 was both below the chance level（25%）.The mean of overall tone perception score was only 36%.There was a confusion matrix where Tone 3 was misidentified more often as Tone 2 than as Tones 1 and 4,and vice versa.Tones 1 and 4 were more often misperceived as Tones 2 and 3.CONCLUSION Formant information（especially for F1）could provide some cues for tone perception,but may not act as an important role.

关 键 词：语音学 汉语声调 线性预测编码技术 共振峰 基频 声调识别 

分 类 号：TN912.3[电子电信—通信与信息系统]