The Development And Equivalent Validation Of Mandarin Monosyllable Words Test List

Speech audiometry is one type of the psychophysical tests, aiming at measuring a person's ability of his auditory system to detect, recognize, identify and understand the speech signals.Two common characteristics of speech audiometry materials are representative and standard, which are the criterions in the selection of test items, the organization of test lists and the acoustic control of the materials. In terms of the development of a speech audiometry material, we should design the test material reasonably according to the test purpose, trying the best to get good reliability, validity and sensitivity, on the base of ensuring the materials'representative and standard.Chinese mandarin speech audiometry began from the efforts of MA Dayou et al. Audiologists like CAI XY, CHENG JY, ZHANG JL, SHEN Y, BAO ZW, GU R, HUANG LN and ZHANG H have developed series of mandarin test materials. However, with the fast development of Audiology in China, early speech audiometry materials can't meet the need of present clinical need. Existing mandarin test materials have some disadvantages, including lack of language update, lack of standardized digital record copies, lack of essential clinical validation, and lack of variety.This paper aims to develop a set of monosyllable audiometry material which will be used to test speech recognition scores, and to explore the process and technique details in making a standard digital mandarin speech audiometry material. With the compromise between the reliability and efficient in clinical practice, we designed 30 test lists with 25 monosyllable test items respectively. During the development, we observed the following principles: succinctness of words, phonemical balance, word familiarity, common words covering, digital recording and processing, and using native speaker.What we have done in this paper were as followed:1) Selected and edited 30 phonemic balanced monosyllable test lists. We calculated the presentation frequency of 22 consonants(including zero consonant), 26 vowels, 4 tunes of Chinese characters in 30 test lists which were composed of 750 monosyllable words, referring to the Chinese phonemes distribution probability summarized in .On the base of PB, we selected 489 monosyllables from the most commonly used 2500 Chinese characters and monosyllable word list in the most commonly used 4000 words, and compiled 30 monosyllable test lists by computer programming and manual adjustment. All the 30 lists observe PB principle, and each list contains 25 monosyllable test items. This set of lists can be used as the fundamental lists of all the Chinese PB monosyllable speech materials.2) Recorded and processed the test material digitally All the 30 lists were recorded digitally by Cool Edit pro 2.0, Creative Sound Blaster card, B&K sound level meter and 4155 transducer in sound insulation booth, and processed by Cool Edit, in which noise reduction, items segmentation, loudness balance were done. Through this stage, we have found out a series of feasible methods to make a good digital speech audiometry material.3) Tested the equivalence of 30 listsEquivalence test for the 30 digitally recorded lists were done in normal-hearing people, and 22 of the lists were selected as formal monosyllable test lists which have feasible characteristics such as good reliability, validity, and time-saving, etc.Equivalence test was done in 2 steps.Firstly, PI curves of each list were draw in order to find appropriate test intensity and adjust each list's difficulty by means of iso-SRT, according to the SRT result of pre-test.Secondly, we implemented a Latin square design to test the equivalence of the lists. Such a design was selected in order to insure enough data be collected and analyzed on each intensity in all the lists. We drew a PI function curve for each list to show the relationship between recognition score and words presentation intensity level, using a logistic regression tool by programming in Matlab. The recognition threshold (dB HL) and PI curve slope (%/dB) were calculated in Matlab.Compared the threshold and PI curve slope of each list using SPSS 12.0 and excluded the lists whose threshold and PI slope varied dramatically. 22 out of 30 lists were remained. Normal people reference recognition threshold of the 22 lists were got.4) Compiled and burnt practical test CD for clinical useWe burnt a practical CD with the tracks of calibration tones and noises, lead sentences, practice lists and test lists in various recording sequences, using Ahead Nero software and SONY R/W CD-ROM. This test CD can meet the need of clinical practice.At the end of this paper, we analyzed and discussed the equivalence test, put forward corresponding amending and modify advices, and point out the research direction of next stage.What's original and pioneer in this paper is that through the before-mentioned work, we got 30 monosyllable mandarin word lists which were following PB principle strictly and 22 of which are equivalent. Also we submitted a set of monosyllable mandarin speech test CD with high quality, which are suitable for testing the speech recognition ability of adults, or children who have a fundamental of more than 2500 monosyllable words. At the same time, through this paper, we have explored the process and technique details in making a standard digital mandarin speech audiometry material, and lay the foundations of the following relative work.