Research Article
Visual Intelligence in Conversational Solutions for Visual Intelligence Security System (VISS)
@INPROCEEDINGS{10.4108/eai.7-12-2021.2314965, author={Ilayaraja N and Madhumitha G and RamVignesh Kumar K and Sitaraman Ramachandrula}, title={Visual Intelligence in Conversational Solutions for Visual Intelligence Security System (VISS)}, proceedings={Proceedings of the First International Conference on Combinatorial and Optimization, ICCAP 2021, December 7-8 2021, Chennai, India}, publisher={EAI}, proceedings_a={ICCAP}, year={2021}, month={12}, keywords={face recognition algorithm tensorflow hu moments hidden markov model vector quantization probability distribution k means clustering}, doi={10.4108/eai.7-12-2021.2314965} }- Ilayaraja N
Madhumitha G
RamVignesh Kumar K
Sitaraman Ramachandrula
Year: 2021
Visual Intelligence in Conversational Solutions for Visual Intelligence Security System (VISS)
ICCAP
EAI
DOI: 10.4108/eai.7-12-2021.2314965
Abstract
Authenticating human beings by Visual features. It is used to improve the accuracy level of Visual Speech Recognition. The Model uses the custom video dataset and pre-processed to grab the sequence of images frames. The process starts by extracting the sequence of images per frame from the video dataset and the lip feature will be extracted from Image. Based on the lip movement the vectors will be detected, integrated and tested with audio features. Facial landmarks such as eyes, nose and mouth region will be extracted. These facial landmarks are used to classify the lip region vector points. These vector points are used for vector quantization and codebook generation. The symbols or codes in the codebook have been considered to determine the accuracy level and whether the user is authenticated. Then, the probability distribution has been determined by considering both Audio and video codebooks. Live dashboards have been developed, which allows authentication of people using pre-trained models.
