REAL-TIME LIP READING USING COMPUTER VISION AND DEEP LEARNING
Diyaneshwaran G
, Hanumanth K S , Jeyanthi A , Nagambika V S
Computer Vision (CV), Deep Learning (DL), 3-Dimensional Convolutional Neural Network (3DCNN), Lip-reading, Region of Interest (RoI).
This project is centered around the development of an advanced speech recognition system designed to accurately identify spoken words from a predetermined lexicon. Employing a sophisticated blend of computer vision and deep learning methodologies, the algorithm is meticulously trained on a sizable dataset meticulously crafted by the project initiators and their collaborators. The dataset encompasses approximately 469 meticulously annotated video clips capturing individuals articulating a diverse range of words, culminating in an extensive 3 GB corpus of training data. The model architecture is carefully crafted, incorporating a sophisticated arrangement of convolutional and dense layers, meticulously orchestrated through the TensorFlow and Keras frameworks.
The training process yields exceptionally promising results, with the model attaining an outstanding 90% accuracy on the training dataset and an impressive 95% accuracy on the validation dataset. These high levels of accuracy underscore the robustness and efficacy of the classification approach. Furthermore, extensive experimentation and optimization efforts ensure the system's reliability and generalization capability, rendering it adept at handling various real-world scenarios with aplomb.
Upon completion of training, the system stands ready to be deployed in live settings, poised to provide accurate and swift word recognition with a remarkable accuracy of 96%. This project represents a significant advancement in the realm of speech recognition, offering a potent tool with far-reaching applications in diverse fields such as human-computer interaction, accessibility, and automation.
"REAL-TIME LIP READING USING COMPUTER VISION AND DEEP LEARNING", JNRID - JOURNAL OF NOVEL RESEARCH AND INNOVATIVE DEVELOPMENT (www.JNRID.org), ISSN:2984-8687, Vol.2, Issue 5, page no.a73-a92, May-2024, Available :https://tijer.org/JNRID/papers/JNRID2405006.pdf
Volume 2
Issue 5,
May-2024
Pages : a73-a92
Paper Reg. ID: JNRID_700244
Published Paper Id: JNRID2405006
Downloads: 000354
Research Area: Science and Technology
Country: Chennai, Tamilnadu, India
ISSN: 2984-8687 | IMPACT FACTOR: 9.57 Calculated By Google Scholar | ESTD YEAR: 2023
An International Scholarly Open Access Journal, Peer-Reviewed, Refereed Journal Impact Factor 9.57 Calculate by Google Scholar and Semantic Scholar | AI-Powered Research Tool, Multidisciplinary, Monthly, Multilanguage Journal Indexing in All Major Database & Metadata, Citation Generator
Publisher: JNRID (IJ Publication) Janvi Wave
