Meu SciELO
Serviços Personalizados
Artigo
Indicadores
- Citado por SciELO
Links relacionados
- Similares em SciELO
Compartilhar
Revista Cubana de Ciencias Informáticas
versão On-line ISSN 2227-1899
Resumo
MONTALVO BEREAU, Ana; REYES DIAZ, Flavio; HERNANDEZ SIERRA, Gabriel e CALVO DE LARA, José Ramón. Spoken language identification for short utterance with transfer learning. Rev cuba cienc informat [online]. 2022, vol.16, n.1, pp. 77-91. Epub 01-Mar-2022. ISSN 2227-1899.
In the present work, spoken language recognition in short utterances was addressed using a convolutional neural network pre-trained on a set of images. Starting from the knowledge transferred from the domain of real images to the audio classification tasks, we assess the impact of multitask learning, taking language recognition as the main task and speaker recognition as auxiliary task. The experiments were carried out on a subset of the Voxforge corpus, and with a significantly lower amount of signals than those used by analog reference systems. The evaluation was done over spectrograms conformed with 3 seconds signal. The results show that the spoken language recognition task benefits from multitasking learning by using the identity of the speaker as an auxiliary task.
Palavras-chave : Spoken language recognition; deep learning; transfer learning; multitask learning..