Automatic Face Recognition Using Legendre Moments and Auto-Associative Memory

Face recognition involves identifying or verifying a person based on their facial features. This study introduces a method using Legendre moments for feature extraction combined with an auto-associative memory approach, specifically a “Multi-connection Hopfield Neural Network” to enhance accuracy and robustness. The method uses a single image per person, optimizing both the computation time and memory usage. Tested on the ORL (Olivetti Research Laboratories) dataset, the approach achieved a 99.0% recognition rate, outperforming other state-of-the-art methods. Additionally, the system demonstrated resilience to noise, maintaining a 95.75 % rate with 50 % random noise.