StrideSense: Faster RCNN for Accurate Human Pathways Gait Recognition

Human gait analysis is quite a multidimensional field, being at the intersection of biomechanics, kinesiology, physiology, computer science, and machine learning, which studies gait (walking) patterns that depend on the interaction of many physiological, social, psychological, and health factors. This is an article on the thorough research about the human gait analysis using the Fast Region-based Convolutional Network (Fast R-CNN) models. This study considers the evolution of a gait analysis from a descriptive to highly advanced tool that has shown its potential for sports, biometrics, security, and healthcare. In this aspect, the research applies deep learning techniques like Convolutional Neural Networks (CNN) to construct the Fast R-CNN gait recognition method that optimally parameters for accuracy and speed. The methodology is collecting data-sets from various sources, preprocessing of the gait images, and then finally training the fine-tuned Fast R-CNN model with the help of transfer learning. Evaluation metrics such as accuracy, recall, and F-score are used to further examine the performance across different gait types, lighting conditions, and background environments. Experimental outcomes show the accuracy, precision, recall, and F1-score at a high level which is a proof that Fast R-CNN algorithm has been successfully used for human gait analysis recognition. The model is capable of achieving a recall rate of 97%, 94% accuracy, 93% precision, and 95% F1-score. These results demonstrate the model can correctly classify broad and diverse human gait variations and report accuracy of over 95%. On top of that, the paper explores where the project heading by incorporating attention mechanisms and recurrent neural networks to improve the model overall and for some real-time deployment.