Avaliação de qualidade vocal com motivação fonética: análise integrada de dados de percepção e acústica

• Autor(a) principal: Silva, Maria Fabiana Bonfim de Lima
• Data de Publicação: 2012
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Biblioteca Digital de Teses e Dissertações da PUC_SP
• Texto Completo: https://tede2.pucsp.br/handle/handle/13573
• topic: Qualidade de voz; Percepção auditiva; Acústica da fala; Medida de produção da fala; Fonética; Voice quality; Auditory perception; Speech acoustics; Speech production measure; Phonetics; CNPQ::LINGUISTICA, LETRAS E ARTES::LINGUISTICA::LINGUISTICA APLICADA

AIM: to verify whether the phonetic nature script Vocal Profile Analysis Scheme can be applied to Brazilian Portuguese (VPAS-BP) or not, for the perceptive evaluation of quality and vocal dynamics of speech samples in a group of teachers with vocal disturbs and laryngeal changes from two public schools in Sorocaba city (SP), as well as to discuss the correspondence with acoustic analysis data. METHODOLOGY: The above mentioned group was formed by 25 female teachers with complaints or voice disturbs and medical diagnosis of laryngeal changes. The stimuli used for the auditoryperceptive analysis refer to records of semi-spontaneous speech (interview and class simulation) and to reading standard texts, totalizing 75 stimuli taken from data banks of previous researches. Those samples were selected on the basis of perceptive experiments (nature and duration of samples) and values in the relationship signal/noise. After the experiments, the samples were edited and analyzed by means of acoustic and perceptive procedures. From the auditory-perceptive point of view, the samples were evaluated by 5 judges; from those, 4 participated in a 15 hour workshop on how to use the VPAS-BP. The 90 samples (75 plus 15 repetitions) were randomly presented through the script Experiment MFC 3.2, applied to the software PRAAT (version 5143). From the acoustic point of view, the same samples were edited through the script SG Expression Evaluator, and it was possible to extract f0, intensity, spectral slope and the long term spectrum (LTS).The perceptive and acoustic data were statistically analyzed: comparison of the average and analysis of multiple correlations (canonic analysis, cluster analysis and discriminant). Features like the agreement and trustfulness inter and intra judges in the VPAS-BP were treated by means of the Cochran and Snedecor test to analyze the behavior of the judges group. It was also made the analysis of the vocal quality profile in the studied samples, the agglomerative cluster hierarchical analysis, the vocal qualities adjustments and vocal dynamics aspects. The acoustic data were analyzed through the agglomerative cluster hierarchical analysis. The procedures of canonical correlation and discriminant analysis were applied to the study of the correlations among perceptive data and acoustic measures. RESULTS: the most frequent adjustments of vocal quality in the studied group were, in decreasing order: laryngeal hyperfunction, harsh voice, lifted larynx, vocal tract hyperfunction, closed jaw, pharynx constriction, lifted tongue body and air escape. The features of vocal dynamics, in decreasing order, were: inadequate breathing support, less variety in pitch, lifted usual pitch, lifted usual loudness, fast elocution rate and raised variety in loudness. The acoustic measures that showed extreme combinations were: mean and absolute deviation of spectral slope and ELT absolute deviation, in one extreme, and in the other, f0 median and quantile of 99,5%. Concerning the correspondence between the perceptive and acoustic data, it was found that the f0 measures (median) and ETL measures (absolute deviation) were close to phonatory adjustments. In one hand, the f0 measures (asymmetry and semi amplitude between quarts) and the intensity (asymmetry) were in the same group of lifted larynx adjustments, less extension of jaw and less variety of loudness, whereas the f0 measures (99,5% quantile) and the spectral slope measures (media, absolute deviation and asymmetry) were in a group with the biggest concentration of adjustments. The most frequent were: laryngeal hyperfunction, closed jaw and lifted tongue body, as well as the dynamic vocal aspects related to inadequate breathing support, lifted usual pitch and loudness, less variety of pitch and loudness. CONCLUSIONS: it was verified that the VPAS-BP was totally applicable to the group of teachers in this study, for the findings of vocal quality and vocal dynamics were compatible with the voice disturbs and laryngeal changes (physiological level) and with the acoustic measures, according to the inclusion criteria in the samples