资源描述:
《噪声下的质性研究》由会员上传分享,免费在线阅读,更多相关内容在工程资料-天天文库。
1、嘈杂语噪声下汉语整句识别的同质性研究【摘要】目的获得嘈杂语噪声背景卜汉语短句识别率■信噪比(Performance-Intensity,P-I)函数的斜率和50%的得分所对应的信噪比(记为SNR50),并对所有短句的同质性进行评估和调整。方法采用16张(每张20句)新编嘈杂语噪声下汉语普通话短句识别表作为测试材料。选取年龄为21〜25岁之间、听力/言语发育正常且FI常以汉语普通话为交流方式的48名受试者按“随机区组设计”在・1、・4、・7、-10dB四种信噪比条件下进行言语识别率的心理声学测屋。使用Statistica7.0软件进行P-I函数拟合和统计分析。
2、结果320句语句在嘈杂语噪声下的言语识别P・l函数曲线的阈值呈正态分布,斜率呈不规则分布。剔除P・l函数强健性增长(斜率>55%/dB)的语句,并保留阈值变异度在±2?滓内的语句,精选出同质性良好的240旬。结论逐个调整每句的SNR以实现同质性,可为编制“等言语识别阈级”的句表奠定扎实的心理声学基础。【关键词】言语测听;嘈杂语噪声;语句;言语识别阈;同质性HomogeneityofMandarinsentenceintelligibilityinbabblenoiseXIXin,ZHAOYang,JIFei,TeresaCHING,HANDong-yiDep
3、artmentofOtolaryngology-HeadandNeckSurgery,ChinesePLAGeneralHospital,Beijing100853,ChinaNationalAcousticLaboratories,126GrevilleStreet,NSW2067,Australia[Abstract]ObjectiveToobtaintheslopeandSRToftheP-IfunctionforMandarinshortsentencesin4・talkerbabblenoise,andevaluate/adjustthehomog
4、eneityallsentencesindividually・MethodsForty-eightMandarin-speakingpeopleagedat18~25withnormalaural/oralcommunicationalabilitieswereparticipatedintheinvestigation.PsychometricmeasuresonMandarinspeechintelligibilityinbabblenoisewereimplementedin-1,-4,-7,-10dBSNRthrough16lists(20sente
5、nceseach)ofMandarinshortsentencesbyaLatin-Squareexperimentaldesign.EstimationandanalysisofP-IfunctionsofallsentencesweredonebyStatistica7.0.ResultsHistogramsaboutthethresholdsandslopesofP-IfunctionsshowedthattheSNR50thresholdswereinGuassiandistribution(?滋±?滓=-6.27±2.04dB),buttheslo
6、pesnot.Twohundredfortysentenceswithhighhomogeneitywerechosenaccordingtotwocriteria:thresholdvariationwaswithin2?滓;slopebelow55%/dB・ConclusionHomogeneityachievedbyadjustmentofSNRforindividualsentencebasedonpsychometricmeasurewouldplayanimportantroleindevelopmentofspeechtestmaterials
7、・【KeyWords]Speechaudiometry;Babblenoise;Sentences;Speechrecognitionthreshold;Homogeneity木课题是开发“嘈杂语噪声下汉语短5识别测听材料”的阶段性工作,该材料以句长6・8个字的普通话儿童短旬作为测试项。统计学上已经阐明,要使测试结果的变异度减小(即具有良好的复测信度),同时又要保证测试的敏感度,应使测试表屮的每一个测试项在可听度(难度)上同质[1]o这就是言语测听材料中有关测试项同质性(Homogeneity)[2]的概念。本课题对320句普通话儿童短句,在不同信噪比条件下
8、进行了言语识別率的心理声学测量,对每个短句的测试结果