Detección de vocales mediante modelado de clusters de fonemas
- Luengo Gil, Iker
- Navas Cordón, Eva
- Sánchez de la Fuente, Jon
- Hernáez Rioja, Inmaculada
ISSN: 1135-5948
Ano de publicación: 2009
Número: 43
Páxinas: 121-128
Tipo: Artigo
Outras publicacións en: Procesamiento del lenguaje natural
Resumo
Many speech signal processing systems require the detection of regions of stability within the signal. As vowels form great stability regions, a system capable of detecting them automatically in the speech is very convenient. This work presents such a system, which uses HMM models of phonetic clusters created according to the acoustic similarities among the phonemes. These clusters are the key element for the system to work correctly in different languages. Although models were trained in Basque, tests were carried out in both Basque and German speech databases showing that the system is able to detect the vowels and their boundaries with acceptable accuracy in both languages.