An Optimized Feature Selection Technique Based on Feature Analysis for Bio-metric Gait Data Classification

For bipedal robots to be stable and successful—especially in human terrain or external environments—optimized feature selection is essential. This work uses statistical techniques to extract important characteristics from bidirectional gait data, which are necessary to accurately simulate human gait patterns. Personalized health evaluation is made possible by the ability to classify irregularities based on a variety of gait data that indicate unique health status. But feature selection is hard, especially when dealing with high-dimensional feature vectors, because handmade methods are insufficient and result in computational inefficiencies. A methodology that first selects features and identifies principal features is presented as a solution to this problem. The best feature set derived from Analysis of Variance (ANOVA) is then utilized by machine learning techniques for data classification. In principal features verification, seventeen features are extracted, which serve as the foundation for the optimized feature set. Three-fold cross-validation is used to assess the model's effectiveness, guaranteeing its resilience and generalizability. This technology can be used to provide bipedal robots with features that enable stable movement and precise human gait imitation. Furthermore, applications in healthcare and rehabilitation may benefit from the capacity to categorize gait data according to health-related factors.