A Robust Hybrid Identification Framework Combines Gait And Face Recognition

It is a common identity recognition approach to utilize the RGB data obtained from cameras. However, most approaches concentrate only on enhancing the recognition performance when complete input data is provided. This paper proposes a hybrid recognition framework that integrates face and gait data to tackle the issue of key data occlusion. Specifically, we introduces DWD, a module designed to integrate facial features and gait features. The IDWA and EDWA submodules allocate dynamic weights to regions within and between the features in DWD, which aligns the facial and gait features and increases the expressive power of the fused features. Experiment shows that our proposed module leads to a marked increase in recognition performance when both input data are complete. Additionally, the model can integrate available relevant data to aid identification even when some or all data is missing.