Lip reading of Sentence with combined model of long short-term memory network and temporal convolutional network.

Résumé

Abstract Lip-reading is the process of recognizing speech based on lip movements. This is a difficult task because the actions of the lips when pronouncing some of the words are identical. Phoneme describes lip motions during a dialogue. During this work, we developed an accurate and robust phoneme sequence recognition system by lip reading. We applied a lip segmentation to input of the proposed network combined of long short-term memory network and temporal convolutional network. Our network is trained to predict phoneme labels from the corresponding mouth area input images. Finally, our evaluation results demonstrate that the use of lip segmented with neural network to predict phrase the performance of our system. Our proposed method improves that accuracy had 61.49% compared to other works.

Citer ce document

Accès au document

Texte intégral en lecture en ligne, réservé aux abonnés SPHAERO et aux membres de l'institution. Se connecter

Voir l'article sur le site de la revue

Auteur(s)

Statistiques

Consultations : 1

Téléchargements : 0