This document presents a study on facial gesture recognition using surface electromyography (EMG) signals and a multiclass support vector machine (SVM) classifier. EMG signals were collected from facial muscles for four expressions: frowning, puckering, smiling, and chewing. Time-domain features like root mean square, variance, and standard deviation were extracted from the EMG signals. These features were classified using SVM and k-nearest neighbors algorithms. SVM achieved the highest accuracy of 92.76% for recognizing the four facial expressions based on the EMG signals. The results indicate this EMG-based approach can efficiently predict different facial expressions.