Full text
| |
| Author(s): |
Fernanda de Barros Vidal
Total Authors: 1
|
| Document type: | Master's Dissertation |
| Press: | Ribeirão Preto. |
| Institution: | Universidade de São Paulo (USP). Faculdade de Filosofia, Ciências e Letras de Ribeirão Preto (PCARP/BC) |
| Defense date: | 2024-04-26 |
| Examining board members: |
Sergio Sheiji Fukusima;
Jamila Viegas Rodrigues;
Aline Epiphanio Wolf
|
| Advisor: | Sergio Sheiji Fukusima |
| Abstract | |
Speech perception is an interdisciplinary topic studied through different approaches. It can be investigated by bimodal (audiovisual) or unimodal (only auditory or visual) association. The visual speech perception, through lipreading, is well explored in several languages, such as English. Although few studies have been carried out on the paradigms that influence this perception in Brazilian Portuguese and no stimulus material with natural faces has been developed and made available considering characteristics that are already well investigated in English. This exploratory research proposed to develop a database of video files for lipreading training in Brazilian Portuguese, through natural stimuli, with words that include different phonemes. 110 videos were produced with two speech therapists pronouncing stimulus words, following guidelines for producing lipreading videos. The stimulus words were selected from the C-Oral Brazil corpora, from Portuguese spoken in informal environments, classified grammatically as nouns, dissyllables and which had the same Consonant-Vowel (CV) structure with the 19 consonant phonemes balanced in the first and second syllable. To test the videos produced, 281 hearing participants aged between 18 and 55 completed a computerized experimental session, hosted on the Pavlovia platform, consisting of 30 videos with the stimulus words. At the end of each video, participants indicated an alternative response, between the stimulus word and two distracting alternatives, which configured possible lexical competitors and varied according to the positioning of the consonant phoneme in the word and received feedback. Descriptive statistics of the data and exploratory analyses were carried out, based on the experimental session. Among the 30 stimulus words, 24 of them had a correct answer rate greater than 70%. The correlation between the average response time and response frequency by response type was calculated using Kendall\'s tau (p<.05). 23 of the 30 items showed a significant negative correlation in relation to correct answers. In incorrect alternatives, the correlation was positive and significant. The proportion of correct responses was higher in words in which at least one of the distracting alternatives differed in terms of manner of articulation. Exploratory cluster analysis was carried out to check possible groupings regarding distinctive features of the phonemes. No grouping pattern by sonority, nasality or manner and place of articulation was observed. However, in general the manner of articulation of plosive had better identification than other phonemes. New research will be able to use the file base (110 videos) produced in this study. The investigation of the visual speech perception in Brazilian Portuguese is at an initial stage. It is vital to consider particularities of the language related to the lexicon and carry out more interdisciplinary research to develop a robust computerized test, understand the relationship between lipreading and cognitive and learning skills, and establish training and research tasks in other areas. (AU) | |
| FAPESP's process: | 22/00801-2 - Speechreading of Brazilian Portuguese: training videos elaboration |
| Grantee: | Fernanda de Barros Vidal |
| Support Opportunities: | Scholarships in Brazil - Master |