Teste para a pesquisa do limiar de reconhecimento de fala: desenvolvimento e estudos psicométricos

Abstract

This research aimed to develop a new test proposal in the Brazilian Portuguese language, in digital recording, specific for the research of Speech Recognition Threshold (SRT), and to search for evidence of validity and reliability for the proposed new test instrument. From a bank of words with evidence of content validity, consisting of 172 dissyllabic words, 19 groups of dissyllables were elaborated, consisting of four words, carefully combined, mainly in quantitative terms of the syllabic structure and of the representativeness of phonemes of the different frequency regions. Next, the 19 groups were analyzed for equivalence, presenting them to 95 normo-listeners, aged 19 and 24, at a signal-to-noise ratio of - 4 dB, and a new test instrument was developed, from the equivalent disyllabic groups, which was organized into two sets of pre-set word groups, to obtain the SRT of each ear separately. Using this new test instrument, the values in dB NA for the SRTs of 20 normal listeners were obtained for analysis of the accuracy of the results obtained according to the tonal thresholds, and of 71 individuals with hearing loss, in addition to the STRs in live voice for this last group, in order to search for evidences of criterion and construct for the new evaluation instrument. In the equivalence study, auditory performance was similar for eight groups (G1, G5, G6, G8, G10, G12, G13 and G19) of the 19 groups of words, with a 50% success score followed by 25%; similar frequencies of scores were not observed for the other groups. For the normal listeners, it was verified compatibility between all the obtained SRTs and the tritonal average of frequencies of 0.5, 1 and 2 kHz, with minimum and maximum difference between these values, of - 5 and 10 dB, respectively. For individuals with hearing loss, most of the SRTs found, in both moments of evaluation, were compatible with the tritonal average of frequencies of 0.5, 1 and 2 kHz. At the first moment of evaluation, the mean value of tritonal average was 58.65 dB NA and the SRTs were 58.90 and 59.41 dB NA, with the new speech material, and the live speech, respectively; and in the retest, the mean value of tritonal average was 57.78 dB NA and the SRTs was 60.17 and 59.33 dB NA. There was a strong correlation, and agreement, between the values of the SRTs obtained from the two evaluation methods. It was concluded that a new evaluation instrument was developed in digital recording, specific for performing the SRT logoaudiometric test, with satisfactory evidence of validity of content, criterion and construct, and satisfactory evidence of reliability.