Smart Monitoring of Leg Exercises in Fitness Centers: A CNN-LSTM Model for Enhanced Gym Safety and Efficiency

Preventing injuries and achieving the exercise's desired effect must be imperative; it applies to exercising in the gym, especially when it comes to the lower body, including squats, lunges, and leg presses with problematic form. This work proposes a CNN-LSTM model used for the appropriate identification of leg-building activities to make real-time exercise tracking and evaluation improved. The model also uses CNN layers to acquire spatial information from frames which are limb locations, and other angles while the LSTM layers study temporal patterns to determine movement pattern over time. Training and assessment yielded reliable performance with cross validation resulted in an accuracy of 94% and precise values of squats 0.94, lunges 0.91 and leg press 0.92. High recall scores of 0.96 for squats, 0.93 for lunges, and, 0.94 for leg press also imply that the application provides extensive coverage of leg workouts that is sensitive to variation. An ablation research carried out on a combination of CNN-only, LSTM-only, and the CNN-LSTM model supported the hybrid architecture with an accuracy rate of 94 percent opposed to CNN-only 88 percent and LSTM-only 85 percent. In addition, ROC analysis provided evidence that the method had a strong discriminative capacity with the three AUC values: 0.97 for squats; 0.95 for lunges; 0.96 for leg presses. Two particular achievements are obvious from the statistical outcomes:. This CNN-LSTM system can be considered as an objective tool for automatic and immediate exercise tracking within the fitness contexts.