Martins, Fernando da AssunçãoLuo, Siyu2020-08-252020-08-252020-07-242020-06-18http://hdl.handle.net/10451/44243O disfarce de voz exacerba a dificuldade do reconhecimento de falante. No entanto, nem sempre as intenções disfarçadas estão presentes. Atualmente, o disfarce inconsciente ainda não é uma área muito investigada, e os estudos relacionados focalizam-se nos disfarces causados pelas mudanças emocionais e pelas razões físicas. O presente trabalho é destinado ao estudo das características acústicas do disfarce inconsciente dos falantes portugueses, do Português Europeu da região de Lisboa (a partir daqui designado como PE), principalmente em ambiente com ruído e com a característica pessoal designada por “sotaque estrangeiro”. Tendo em conta que o perfil do falante é constituído por um conjunto diversificado de parâmetros, o estudo centra-se os parâmetros acústicos utilizados comummente em aplicações de análise acústica, bem como os mais referenciados na literatura - fundamentalmente nos parâmetros Frequência Fundamental, Formantes, VOT, Jitter, Shimmer e HNR. Com este propósito, foram gravados doze indivíduos, incluindo seis informantes nativos e seis não-nativos, que leram um conjunto de frases nas seguintes condições de produção: voz normal, voz disfarçada em ambiente com ruído, voz disfarçada com sotaque estrangeiro. Na primeira simulação de disfarce, todos os informantes falam com auscultadores de ouvido, reproduzindo uma mesma sequência de barulho de lugares públicos. Na segunda simulação de disfarce, os informantes não nativos produzem as sequências de fala do corpus naturalmente, sem ruído. Primeiro, os informantes não-nativos não foram informados sobre o objetivo de estudo, de modo a que mantenham o sotaque estrangeiro original e de que não disfarcem as suas vozes conscientemente, e depois, foram informados produzir as frases com a maneira mais aproximadas dos informantes nativos. As características acústicas do disfarce sem consciência, com consciência e sem disfarce são aqui estudadas de uma forma contrastiva, usando o software Praat. Por um lado, verificar a eficácia dos parâmetros mencionados na distinção de falantes e, por outro lado, observar a influência dos disfarces mencionados nos vários parâmetros acústicos. Os resultados obtidos demonstram que: (i) em termos de disfarce de sotaque estrangeiro, as influências da língua materna envolvem quase todos os parâmetros deste estudo. No aspeto de validade no reconhecimento de falante, os triângulos vocálicos, o tempo de duração das vogais e o VOT parecem ser os parâmetros acústicos mais potentes do que outros; (ii) em termos do disfarce com ruído, os parâmetros acústicos mais eficazes no reconhecimento de falantes parecem ser o F0, o F1, o tempo de duração das vogais, o VOT, e o HNR; (iii) depois do período de aquisição do idioma, embora os falantes não-nativos tentem alterar a sua pronúncia intencionalmente para ser semelhante à de falantes nativos, o efeito de disfarce não foi totalmente conseguido. Em suma, o estudo presente pretende-se: (i) fornecer informações para os falantes chineses na aquisição de segundo idioma – Português Europeu; (ii) complementar as lacunas da área de estudo de disfarce inconsciente; (iii) contribuir para inferir o ambiente em que os falantes estão localizados e a identidade de identificação forense de falantes em situações judiciais.Voice disguise increases the difficulty in the recognition of speakers. However, disguised intentions do not always exist. Currently, researches on unconscious guise are still rare. Related research focuses on disguise caused by emotional changes and physical causes. The present study is aimed at investigating the acoustic characteristics of the unconscious disguise of European Portuguese speakers from the Lisbon region (from now on, PE), mainly in a noisy environment and with the personal characteristic called “foreign accent”. Considering the speaker's personal profile consists multiple different parameters, only the acoustic parameters commonly used in acoustic analysis applications or most referenced in the literature such as fundamental frequency, formants, acoustic triangle, vowel duration, VOT, and voice quality parameters: Jitter, Shimmer and HNR will be studied. For this purpose, twelve people were recorded, including six locals and six non-local informants. They read a set of phrases under the following conditions: normal speech, disguised sounds in noisy environments, disguised sounds with foreign accents. In the first disguise simulation, all speakers speak wearing headphones which will play a reproduced noise sequence from public places. In the second simulation, non-native speakers produce the speech sequences of the corpus naturally without any noise. Speakers were not informed about the purpose of the study, so that they maintain the original foreign accent and without consciously disguising their voices. Then, they were told to speak in a way closest to the native speakers. Here, the acoustic properties of unconscious disguise, conscious disguise and no disguise are studied using software Praat in a comparative manner to verify the validity of the parameters mentioned above when distinguishing speakers and observe the effects of the above mentioned disguise on various acoustic parameters. The obtained results show that: (i) in terms of the disguise of foreign accents, the influences of the mother tongue involve almost all the parameters of this study. In the aspect of validity in speaker recognition, the vowel triangles, the duration of the vowels and the VOT seem to be the most potent acoustic parameters compared with others; (ii) in terms of the disguise of noise environment, the most effective acoustic parameters in recognizing speakers appear to be F0, F1, vowel duration, VOT, and HNR; (iii) after the language acquisition period, although non-native speakers intentionally try to change their pronunciation to be similar to that of native speakers, the disguise intention has not been fully achieved. In a word, this study aims to: (i) provide the second language (European Portuguese) learning information for Chinese speakers; (ii) fill in the gaps in the field of unconscious disguise research; (iii) help infer the circumstances of the speaker and then technically support the identity of the speaker’s judicial appraisal under legal circumstances.porDisfarce não consciente em Fonética Forense – a Voz Femininamaster thesis202510034