A MATHEMATICAL METHOD TO VALIDATE NEW ROMANIAN SPEECH AUDIOMETRY MATERIALS FOR EVALUATION THE HEARING LEVEL IN YOUNG AND ADULTS Cover Image

A MATHEMATICAL METHOD TO VALIDATE NEW ROMANIAN SPEECH AUDIOMETRY MATERIALS FOR EVALUATION THE HEARING LEVEL IN YOUNG AND ADULTS
A MATHEMATICAL METHOD TO VALIDATE NEW ROMANIAN SPEECH AUDIOMETRY MATERIALS FOR EVALUATION THE HEARING LEVEL IN YOUNG AND ADULTS

Author(s): Cristina Gena DASCALU, Sebastian Cozma, Gabriel DIMITRIU, Mihaela MOSCALU, Raluca Olariu
Subject(s): Social Sciences
Published by: Carol I National Defence University Publishing House
Keywords: speech audiometry; communication abilities; Romanian language; fitting test; mathematical validation.

Summary/Abstract: The hearing assessment is an important step in detecting any possible hearing loss which leads to specific communication and speech pathologies. Different speech intelligibility tests have been developed for some languages, but calibrated materials for Romanian language were not validated until now. We used a new originally kit test containing monosyllabic and bisyllabic words from spoken language, corresponding to the normal speech. The words were filtered by special software in order to select the most common and frequency spectrum homogeneous word lists, and the audiological assessment was made in standard test conditions. The aim of the study was to mathematically assess and validate this material in Romanian language for speech audiometry. In order to do this we recorded for each list the percentage of words correctly identified by different subjects with normal hearing at different sound intensities (in dB) and we generated the corresponding curve, which was then compared with the ideal curve known from literature, corresponding to a normal hearing. The difference between these curves was calculated as a normalized root mean squared error (NRMSE), and a tested list was validated when this difference was small enough. The simulations were made in MATLAB, and we wrote a dedicated algorithm in order to generate the curves and to calculate the proposed error. We tested 260 lists, among them 85.4% were identified with a NRMSE error smaller than 15%; these lists were kept as valid, the others being reformulated. The behaviour of the valid lists was further statistically analysed, in order to prove its usefulness as a clinical tool for widely use in the evaluation of hearing loss or communication disorders.

  • Issue Year: 13/2017
  • Issue No: 01
  • Page Range: 522-528
  • Page Count: 7
  • Language: English