This research investigates the potential of electromyography (EMG) signals to understand a user’s intention for upper-limb rehabilitation. We focused on classifying elbow flexion and extension movements using EMG data collected from healthy participants. By analyzing EMG signals alongside elbow joint angle, torque, and velocity, we identified a time delay between when a user intends to move and when the movement actually begins. The study also highlighted the influence of individual differences in how EMG signals manifest, emphasizing the need for robust classification techniques that can account for this variability. These findings provide a valuable foundation for developing EMG-based control systems for exoskeletons used in rehabilitation. Future work will explore more advanced classification algorithms, incorporate a normalization technique to account for variations in muscle strength across users, and expand the study to include individuals with neurological conditions. This broader investigation will be crucial for developing control systems that can be applied in real-world rehabilitation settings for a wider range of patients.