谷歌浏览器插件
订阅小程序
在清言上使用

The Development and Evaluation of the Finnish Matrix Sentence Test for Speech Intelligibility Assessment.

Acta oto-laryngologica(2014)

引用 31|浏览16
暂无评分
摘要
Conclusion: The Finnish Matrix Test is the first sentence test in noise for the Finnish language. It was developed according to the HearCom standards and provides reliable speech intelligibility measurements with highly comparable results with the other international matrix tests. Objectives: The aim of the study was to develop an accurate speech intelligibility test in noise for the Finnish language that is comparable across different languages. Methods: We chose a matrix sentence test, which comprises a base matrix of 10 names, verbs, numerals, adjectives and nouns. Test lists were formed from this matrix quasi randomly, providing test sentences of the same syntactical structure. The speech material corresponds to everyday spoken language and the phoneme distribution is representative of the Finnish language. The test was optimized by determining the speech recognition thresholds of the individual words and subsequently by applying level corrections of up to +/- 3 dB. Evaluation measurements were performed to check the equivalence of the different test lists with respect to speech intelligibility and to provide reference values for further clinical applications. Results: After training, the mean speech recognition threshold (SRT) and the slope of the final test lists were - 10.1 +/- 0.1 dB signal-to- noise- ratio (SNR) and 16.7 +/- 1.2%/ dB, respectively (measurements at constant level; inter-list variability). The mean SRT and the slope of the test subjects were - 10.1 +/- 0.7 dB SNR and 17.5 +/- 2.2%/dB (measurements at constant level; inter-subject variability). The expected SRT range for normal-hearing young adults for adaptive measurements is - 9.7 +/- 0.7 dB SNR.
更多
查看译文
关键词
Speech audiometry,speech recognition in noise,signal-to-noise-ratio (SNR),speech recognition threshold,SRT,matrix test
AI 理解论文
溯源树
样例
生成溯源树,研究论文发展脉络
Chat Paper
正在生成论文摘要