Development and evaluation of a disyllabic Mandarin word discrimination test

• Main Authors: Yao-Yu Chiang, 蔣燿宇
• Other Authors: Shuenn-Tsong Young
• Format: Others
• Language: zh-TW
• Published: 2005
• Online Access: http://ndltd.ncl.edu.tw/handle/43919500223559298472

碩士 === 國立陽明大學 === 醫學工程研究所 === 93 === The most important sounds transuded by the ear are sounds that compose speech signals. Speech itself is the most appropriate material to explore the hearing capability. Most current speech test are designed for English speaking population. However, there are more than 90% of people who do not speak English and there is no standardized speech test for them. Mandarin is different from English in the aspect of tones which are associated with different meanings. There are only a few studies focusing on Mandarin word discrimination test (WDT). Therefore, the purpose of this study is to develop a word discrimination test with disyllabic Mandarin word for assessing the ability to perceive speech in general communication conditions. For a word discrimination test, the test materials usually have to follow three criteria: familiarity, phonemic balance and homogeneity. These criteria make the selection of test materials become a complex issue. Genetic algorithm performs an efficient, robust, and low-complexity search of the problem space, so this study used this method to select test materials to match these criteria. Test materials were evaluated by certain homogeneous experiment to prove them to to match the criteria. The result show that the test materials are highly similar based on performance norm analysis. In addition, performance of every test materials in norm has the standard deviation of 5dB even at different presentation levels. There are three significant contributions in this study. First, this study provided six disyllabic Mandarin word lists which were evaluated by homogeneous experiment, and they can be applied in clinical application. Secondly, this study constructed a homogeneity experiment platform and a word discrimination test platform with disyllabic Mandarin words. Both platforms can be automatically conducted. Finally, this study created a standard procedure to produce test materials which are homogeneous. This procedure includes several steps which were: 1) build up a corpus database; 2) estimate the target factor according to the contributions of vowels, consonants and tones; 3) create testing materials by using genetic algorism; and 4) finally, conduct the homogeneous experiment. Therefore, in the future, test materials for different purposes can also develop by using this standard procedure.