普通话快速噪声下言语测听材料的等价性评估  

Equivalence study of mandarin quick speech in noise test materials

作　　者：任丹丹[1] 张华[1] 武文芳[1] 曹文[2] 郭连生[1] 陈建勇[3,4] 李玉玲[1] 王硕[1] 吴丹[1] 

机构地区：[1]首都医科大学附属北京同仁医院耳鼻咽喉头颈外科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),北京100730 [2]北京语言大学对外汉语研究中心,北京100083 [3]上海交通大学医学院附属新华医院耳鼻咽喉头颈外科 [4]上海交通大学耳科学研究所,上海200092

出　　处：《中国耳鼻咽喉头颈外科》2012年第3期137-141,共5页Chinese Archives of Otolaryngology-Head and Neck Surgery

基　　金：国家自然科学基金资助项目(81070784)

摘　　要：目的开发一套标准的普通话版快速噪声下言语测听材料,评估依据听力学要求和语言学特征选取的300个语句材料的等价性,并获得其在50%正确识别时所对应的信噪比(记为SNR-50)的正常值。方法首先将这些语句材料按照数字化录制原则录制为女声音频材料,背景噪声为4人同时谈话的言语噪声。言语声和噪声混合在一起。选取30例正常受试者,采用听说复述方式,分别在+6、+3、0、-3、-6 dB信噪比条件下测试。对测试所得结果使用SPSS17.0统计软件进行数据分析和强度-识别率函数曲线的拟合。结果 300个语句在噪声环境下测试结果分析显示,斜率和阈值不呈正态分布。剔除斜率过大(>0.72)的语句,及曲线拟合相关系数过小(r<0.7)的语句,精选出等价性较好的78句。后者呈正态分布。结论经过对300个语句的等价性评估,经过严格的筛选及精细调整,最终保留的78个语句在正常人中的等价性较好,下一步可进行这些语句复测信度的测试。OBJECTIVE To select equivalent sentences and get their SNR-50 value,in order to develop the Mandarin Quick Speech in Noise Test（M-Quick SIN）.METHODS All the 300 sentences were recorded into the digital materials with a female voice at first,the background noise was 4-talker babble noise,and they were mixed together.Then they were presented to 30 subjects at ＋6,＋3,0,-3,-6 dB SNR.The subjects should repeat the sentences that they heard.Evaluation and analysis of P-I functions of all sentences were done by SPSS 17.0.RESULTS The Histogram and the analysis showed that neither the thresholds nor the slopes were in the normal distribution.The sentences,whose regression coefficients were below 0.7 and slopes were greater than 0.72 were abandoned.The retained 78 sentences had good equivalence.CONCLUSION Through the evaluation and selection,the retained 78 sentences,which were adjusted for their SNR（signal-to-noise ratio）,had good equivalence among normal hearing people.Next,test-retest reliability should be done for these sentences.

关 键 词：语言试验 言语识别测验 噪声 语句材料 等价性 

分 类 号：R764[医药卫生—耳鼻咽喉科]