Soham Akhade, Prof.J.C.Musale, Prof.S.J.Nawale, Omkar Jadhav, Atharva Bhadale, Prathmesh Gaikwad

Visual Speech Recognition( VSR) is a fleetly evolving field with different operations in mortal- computer commerce, availability, and security. This paper presents an innovative approach to VSR, fastening on the birth and analysis of lip movements for speech recognition. Traditional speech recognition systems calculate primarily on aural information, making them vulnerable to noisy surroundings and audio disturbances. In discrepancy, our proposed system leverages the visual modality by employing the rich information decoded in lip movements during speech product The study begins by collecting a comprehensive data-set of visual and audio recordings of speech in colorful languages and surrounds. latterly, a deep literacy armature is designed to process the visual data, emphasizing lip movements, and the corresponding audio data. The proposed model integrates conventional neural networks( CNN s) and intermittent neural networks( RNNs) to prize and fuse information from both modalities. This emulsion process enhances the robustness of the system by mollifying the limitations of traditional audio- only speech recognition We estimate the performance of the visualbased speech recognition system on a range of standard datasets and real- world scripts. The results demonstrate the efficacity of our approach, pressing its capacity to ameliorate recognition delicacy, particularly in noisy surroundings or situations where audio data is deficient or unapproachable In conclusion, our exploration contributes to the advancement of Visual Speech Recognition by introducing a new approach that emphasizes lip movement analysis. By using both audio and visual modalities, the proposed system provides a more robust and protean result for speech recognition, with the implicit to enhance operations in mortal- computer commerce, availability, and security.

https://doi.org/10.58357/jedcs.2024.v3.i1.5