GaitAI: Cutting-Edge Machine Learning for Biometric Gait Recognition and Analysis
Authors:-M. V. Rajesh, Putsala Pujitha, Pothula Mohana Surya Kumari, Guttula Naveen Sagar, Veesam Vamsi, Dara Prudhvi Narayana
Abstract-This study offers a comprehensive investigation into the field of gait recognition in biometric analysis, focusing on the specific challenges associated with using gait as a biometric feature. The research evaluates various machine learning (ML) algorithms, including Individual Node Evaluation, Statistical Inference, Regression Modelling, Support Vector Machines, Nearest Neighbour Classification, Decision Tree Structures, Random Forest Ensembles, and Multilayered Neural Networks. Thorough testing is conducted to assess the performance of each model in accurately identifying individuals based on their unique gait characteristics. The approach emphasizes extensive preprocessing to maintain data quality and relevance. Additionally, Sequential Backward Selection (SBS) is employed for feature selection, along with dimensionality reduction techniques such as Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA), which improve the model’s focus on key features. The research also investigates deep learning models, analysing different architectures to assess their effect on gait recognition accuracy. A detailed comparative analysis evaluates the advantages and limitations of each method, providing valuable insights for the field. By exploring a variety of ML and DL approaches, this study sets a benchmark for future developments in biometric security. It highlights the potential of gait recognition as a reliable, non-invasive identification method, paving the way for the creation of more advanced and precise biometric systems that are crucial for the evolving needs of security and personal identification.
