Loading...
Publication
Parâmetros acústicos para a identificação do falante em casos de disfarce de voz eletrónico deliberado
| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 4.19 MB | Adobe PDF |
Authors
Advisor(s)
Abstract(s)
A presente investigação teve como objetivo estudar a relevância dos parâmetros acústicos da fala na identificação do falante em contexto de disfarce de voz eletrónico deliberado. Pretendeu-se verificar a robustez destes parâmetros, bem como identificar quais sofrem alterações e quais se preservam, tendo como base comparativa a voz normal. Os parâmetros analisados foram: proporção harmónico-ruído (HNR), jitter, shimmer, frequência fundamental (F0), formantes das vogais orais tónicas (F1 e F2), bem como a oclusão, a explosão e o tempo de início de vozeamento (VOT) das consoantes oclusivas não vozeadas.
Participaram 10 informantes – cinco do sexo feminino e cinco do sexo masculino – que leram um texto foneticamente equilibrado na condição de voz normal. As gravações foram posteriormente manipuladas sob duas condições de disfarce eletrónico: +6 semitons e -6 semitons. A voz normal foi utilizada como referência para a análise comparativa com as vozes manipuladas, permitindo observar a influência dos disfarces sobre os diferentes parâmetros acústicos.
This study aims to investigate the relevance of speech acoustic parameters in the identification of speakers under deliberate electronic voice disguise. It seeks to assess the robustness of these parameters and determine which ones are affected or preserved, using the normal voice as a reference. The parameters analysed were harmonics-to-noise ratio (HNR), jitter, shimmer, fundamental frequency (F0), formants of oral tonic vowels (F1 and F2), as well as occlusion, burst, and voice onset time (VOT) of voiceless stop consonants. Ten participants – five female and five male – read a phonetically balanced text in the normal voice condition. The recordings were then digitally manipulated under two voice disguise conditions: +6 semitones and –6 semitones. The normal voice served as a baseline for comparing the manipulated recordings and for assessing the impact of voice disguise on these acoustic parameters. The results showed that all parameters underwent alterations under the disguise conditions, which hinders speaker identification. Nevertheless, in some cases, these changes provided acoustic cues that may assist in excluding certain suspects.
This study aims to investigate the relevance of speech acoustic parameters in the identification of speakers under deliberate electronic voice disguise. It seeks to assess the robustness of these parameters and determine which ones are affected or preserved, using the normal voice as a reference. The parameters analysed were harmonics-to-noise ratio (HNR), jitter, shimmer, fundamental frequency (F0), formants of oral tonic vowels (F1 and F2), as well as occlusion, burst, and voice onset time (VOT) of voiceless stop consonants. Ten participants – five female and five male – read a phonetically balanced text in the normal voice condition. The recordings were then digitally manipulated under two voice disguise conditions: +6 semitones and –6 semitones. The normal voice served as a baseline for comparing the manipulated recordings and for assessing the impact of voice disguise on these acoustic parameters. The results showed that all parameters underwent alterations under the disguise conditions, which hinders speaker identification. Nevertheless, in some cases, these changes provided acoustic cues that may assist in excluding certain suspects.