Fusion of CNN and LSTM Models for Enhanced Gait Swing Phase Identification

This paper presents a CNN-LSTM hybrid model intended for precise recognition of gait swing phases, explicitly focusing on initial swing, mid swing, and terminal swing. Gait analysis plays an important part in different fields including medical care and biomechanics, offering experiences in motion functions and supporting rehabilitation techniques. The proposed model incorporates a innovative approach of Convolutional neural networks (CNNs) for feature extraction from gait successions and long short-term Memory networks (LSTMs) for catching temporal conditions inside each swing stage. After processing the sequential gait data, the CNN component extracts hierarchical features that emphasize spatial patterns important for differentiating between swing phases. These features are then used to feed into layers of Long Short-Term Memory (LSTM) models, which simulate the temporal dynamics present in human gait and allow the system to accurately identify small phase changes. Performance parameters such as accuracy, precision, recall, and F1-score are used to evaluate the model's performance and show how well it can categorize inclusion gait swing phases. The datasets were tested on different models CNN-LSTM, CNN-Dense-LSTM, and LSTM and the results show that CNN-LSTM achieved the highest accuracy of 98.56%.