Modified Rhyme Test for Evaluating Turkish Speech Intelligibility

Umut Arıöz ,Banu Günel

doi:10.12996/gmj.2015.25

Umut Arıöz , Banu Günel

Open Access

https://doi.org/10.12996/gmj.2015.25

Copy DOI

Abstract

Objective: There are both objective and subjective tests for evaluating the speech intelligibility. Speech Intelligibility Index (SII) and Modified Rhyme Test (MRT) are the most common methods for objective and subjective tests, respectively. While objective tests do not depend on the language, subjective tests are applied to the hearing-impaired subjects in their language. In these tests, hearing-impaired subjects try to recognize the sound by looking at the word lists in their hand. The word lists can be presented as two-word or six-word choices in design. However, the preferred type a is six-word choice for MRT. Up to now, these word lists were developed in different languages with different methods. For Turkish language, only two-word choice lists were developed; there was no study related to the Turkish language with six-word choice. In this study, we aimed to develop applicable six-word Turkish MRT lists by using hearing-impaired subjects. Methods: In this study, three different six-word Turkish MRT lists were developed according to the phonetic characteristics of the Turkish language, and these lists were tested via 12 hearing-impaired subjects. In the lists, 5 different Turkish phonetic characteristics were taken into account. For the analysis, the detailed audiograms (18 octave band) of hearing-impaired subjects were used. Results: Detailed statistical analysis for phonetic variations were shown for each hearing-impaired subject. Having correct structure and robust word choices for MRT lists were shown by comparing the equivalent speech spectrum levels of sounds and MRT results of the hearing-impaired subjects. Conclusion: New Turkish MRT lists were developed according to the phonetic characteristics of the Turkish language. Usability of the lists was shown by audiological test with the hearing-impaired subjects. Statistical analysis and the results of tests showed that developed Turkish MRT lists can be used in audiological studies as a reference material.

Highlights

Speech intelligibility measures are used mainly for evaluating synthetic or processed speech
The equivalent speech spectrum levels of 23 fas pas kas yas tas bas sustention recordings, box-plot results, and percentages of correct choices of modified rhyme test (MRT) for each phonetic group were calculated for all subjects
Equivalent Speech Spectrum Levels (ESSLs) equivalent speech spectrum level (ESSL) is calculated by taking the logarithms of root mean squares of each 28 jul bul çul dul kul pul sustention frequency band

Summary

Introduction

Speech intelligibility measures are used mainly for evaluating synthetic or processed speech. Text to speech (TTS) synthesis systems and hearing loss simulation (HLS) are examples of the main fields for speech intelligibility measures [1,2,3]. TTS systems try to convert normal language text into speech. These systems convert the text into symbols (the text normalization part) and, after applying linguistic analysis to them, convert these symbolic representations into sounds. The primary goal of HLS studies is to model the impaired hearing of individuals, thereby eliminating the need to involve subjects with hearing loss in listening tests. For evaluating the effectiveness of the HLS, both subjective and objective speech intelligibility measures exist. The most widely used subjective speech intelligibility measures are the diagnostic rhyme test (DRT) [7], modified rhyme test (MRT) [2] and mean opinion score (MOS) [8]

Methods

Results

Conclusion