Enhanced Vitality Detection in Multiple Human Facial Images using Encoding Convolutional Neural Networks for Superior Accuracy and Reliability

Enhanced vitality detection in human facial images is essential in fields like healthcare and security, especially with varying image conditions. This study presents a novel method for detecting vitality by integrating the Swin Transformer, Large Separable Kernels (LSK), and Newton-Raphson-based Optimizer (NRBO), resulting in a marked enhancement in the accuracy of detecting subtle signs of vitality. This unified framework forms an Encoding Convolutional Neural Network (ECNN), called Swin Large Separable Newton Convolutional Kernel (SLNCK), which encodes multi-scale and multi-domain facial information into optimized feature representations. SLNCK provides an end-to-end architecture that bridges attention-based learning, efficient convolution, and rapid optimization for robust facial vitality detection. The approach effectively encodes facial features, detecting physiological and emotional indicators such as micro-expressions and skin tone variations in dynamic, multi-angle data. By integrating Swin Transformers, the model captures long-range dependencies and local features, enhancing spatial awareness. LSK optimizes convolution, reducing computational complexity while maintaining high feature resolution, vital for accurate assessment. This model offers superior performance in detecting subtle vitality cues, benefiting healthcare, security, and assistive technologies. The proposed method demonstrates superior efficiency compared to existing methods and delivers impressive results, with an accuracy of 96.7%, a precision of 97.9%, and an F1 score of 97.3% these results highlight its effectiveness in delivering more accurate and reliable performance, surpassing the performance of current techniques.