摘要:
用普通话若干孤立单韵母作语音材料,有三级性别或年龄不同的发音人,通过声学分析和统计分析,比较了言语识别预加工的三种听觉模型.这些模型由几种不同的听觉变换和规范构成.实验的效果表明,最佳的听觉变换为Erb,但几种规范之间却没有明显的差别.Chistovich的频谱重心效应对普通话单韵母也是适用的,韵母的临界距离估算为3.0Bark或3.5Erb.
Abstract:
As experimental phonetic materials,this paper uses some standard Chinese monofinals produced isolatedly by three groups of speakers of different sex or age.By means of acoustical and statistical analysis,three auditory preprocessing models of speech recognition,that contain some kinds of both auditory transformations and normalizations,are compared.Experimental resultsshow that Erb is the best one among auditory transformations,but no significant difference is found among these normalizations.Chistovich's SCG effect is verified,the critical distance between standard Chinese finals is estimated as 3.0 Bark or 3.5 Erb.
下载:
