Genetic Algorithm-Based 3D Feature Selection for Enhanced Lip Reading Accuracy

Lip reading is a critical technology in speech recognition, assistive communication, and human-computer interaction.Traditional methods often rely on 2D features, which fail to capture the depth and complexity of lip movements, leading to reduced accuracy.This project proposes a Genetic Algorithm-Based 3D Feature Selection framework to enhance lip reading accuracy by leveraging 3D spatial features that provide a richer representation of lip dynamics.The high dimensionality of 3D features can introduce redundancy and noise, which may hinder model performance.To address this, a Genetic Algorithm (GA) is employed to optimize feature selection, ensuring only the most relevant features are used for training.The GA iteratively selects and evaluates feature subsets based on their impact on model accuracy, reducing computational overhead while improving performance.Experimental results demonstrate that the proposed system outperforms traditional 2D-based methods, achieving higher accuracy, precision, and efficiency.This approach highlights the effectiveness of combining 3D feature extraction with genetic optimization, offering a scalable solution for more accurate and robust lip reading systems.