Gait-Based Person Identification with Group-First Model Filtering

The identification system is segmented into three distinct stages using three different recognition models: a first model for group determination, a second model for initial identity assessment, and a third model for precise identity determination. Each model handles a specific aspect of the identification task, allowing the system to achieve high accuracy by dividing the complex identification process into manageable segments with specialized models.

The system performs preliminary group determination using the first gait recognition model before proceeding to identity-specific analysis with the second and third models. This preliminary action filters out non-members early in the process, reducing the computational burden on subsequent models by only processing data from relevant groups, thereby optimizing resource utilization while maintaining high recognition precision.

The patent employs gait datasets as intermediaries that capture movement patterns without requiring direct physical contact between the authentication device and the user. These datasets serve as mediators that preserve the dynamic biometric features necessary for accurate identification while maintaining the contactless operation mode, thus resolving the contradiction between convenience and measurement precision.