语言清晰度试验中数据有效性检验方法的研究

An approach to validity test of speech articulation testing data

语言清晰度试验是评价语言交流场所音质优劣最直接的方法,而试验数据有效性检验是语言清晰度得分统计平均的基础。本文采用标准差检验、邻差型检验和半极差型检验这三种方法对语言清晰度试验数据进行了有效性检验,并对异常数据的处理也进行了讨论。结果表明对于语言清晰度试验数据的有效性,邻差型检验或半极差检验优于标准差检验,而以邻值替代异常数据则更接近实验原始数据。

Speech articulation test is the most direct way to do acoustics evaluation in places for speech communication. Average speech articulation scores can be obtained from the statistical analysis results based on the validity test of the experimental data. The validity tests of speech articulation experimental data are made with the methods of standard deviation test successively, difference test from adjacent data and extreme-mean difference test successively. Abnormal data processing is also discussed. It is shown that the diff-erence test from adjacent data and the extreme-mean difference test are better than the standard devi-ation test for the validity test of the speech articulation experimental data. And the processing samples are closer to the initial experimental data when abnormal data are substituted by adjacent values.