Enhancing Gait Recognition with a CNN-RF Approach: Focus on Single and Double Support Phases

Gait recognition has been the subject of much research, and deep learning techniques have supported this field of study. This research uses Random Forest (RF) and Convolutional Neural Networks (CNN) to provide a unique method for identifying gait phases, namely the stance and swing phases. With the use of a Kaggle dataset containing 8227 normalized images from five different people, this research attempts to correctly categorize gait phases according to the ideas of single support and double support. The hybrid model CNN+RF is using. Convolution neural networks were utilized to extract features, and the output was fed into an RF classifier to carry out the classification. Our findings show the gait analysis potential of these machine learning algorithms, with implications for biometric identification, sports science, and rehabilitation applications. The results add to the body of knowledge on human movement recognition and point to directions for future study into improving gait analysis systems. From the results, it was clear that the highest accuracy was achieved in the single support stance phase which is $\mathbf{9 0. 7 0 \%}$ and the overall accuracy achieved by the model was about $\mathbf{8 5. 9 \%}$ and gives a social security.