The document discusses a proposed multimodal biometric system that uses feature-level fusion of fingerprints, face, and eye tracking biometrics. It extracts features from each biometric individually and then uses a joint sparse representation method to fuse the features together while accounting for noise and occlusion. This sparse representation forcing the different biometric features to interact through shared sparse coefficients. The proposed system aims to make multimodal biometrics recognition more robust to problems like noisy data, non-universality of single biometrics, intraclass variations, and spoof attacks.