This study analyzes the symmetry of facial expressions using surface electromyography (sEMG) to detect facial muscle activity and proposes a novel method to assist in diagnosing facial paralysis. The experiment employed the Trigno device by Delsys for EMG signal acquisition, with seven sets of facial expression movements designed to capture sEMG signals. Data extraction and processing were performed on the MATLAB platform using coarse data extraction and energy threshold algorithms. Three key features—Mean Absolute Value (MAV), Waveform Length (WL), and Root Mean Square (RMS)—were extracted from the six-channel sEMG signals. The actions were classified using a Support Vector Machine (SVM) based on these features, and combined with a probability distribution model, automatic recognition and symmetry calculation of the facial actions were carried out. The symmetry standard for different actions at a specified probability was determined using the biased POL confidence interval. The results indicate that this method accurately recognizes facial actions and their symmetry, showing promising potential for clinical applications.