Silent Speech Recognition

Speech is essential to exchange information. Speech recognition is one of the interfaces for man-machine interaction. However, the performance of these systems is restricted to noisy acoustic conditions. Silent speech i.e. visual dynamic features of speech have more potential information for Human-C...

Full description

Saved in:
Bibliographic Details
Published inCognitive Computing and Information Processing Vol. 801; pp. 130 - 139
Main Authors Kandagal, Amaresh P., Udayashankara, V., Anusuya, M. A.
Format Book Chapter
LanguageEnglish
Published Singapore Springer Singapore Pte. Limited 2018
Springer Singapore
SeriesCommunications in Computer and Information Science
Subjects
HMM
Lip reading
Otsu
Speech recognition
Online AccessGet full text

Cover

More Information
Summary:Speech is essential to exchange information. Speech recognition is one of the interfaces for man-machine interaction. However, the performance of these systems is restricted to noisy acoustic conditions. Silent speech i.e. visual dynamic features of speech have more potential information for Human-Computer Interaction. This paper presents lip localization and segmentation by Otsu algorithm. The height and width parameters of lip movements are captured as visual cues for silent speech recognition. We develop stochastic visual word models with an in-house database of 20 subjects. Performance evaluation these models are measured by word error rate. The accuracy of the system recorded for speaker dependent female subjects is 84.6%, and 65.8% as an overall result.
ISBN:9789811090585
9811090580
ISSN:1865-0929
1865-0937
DOI:10.1007/978-981-10-9059-2_13