Fatigue of the muscles involved in swallowing is regarded as one of the factors that increase the risk of aspiration and choking. However, an objective method for assessing muscle fatigue in swallowing-related muscles during the act of swallowing, which consists of voluntary and involuntary reflexive movements and lasts about 1 s, has yet to be established. Therefore, we aimed to develop a method to noninvasively evaluate muscle fatigue from surface electromyography (sEMG) signals recorded during swallowing. In 12 younger and 11 older healthy adults, 44-channel sEMG signals were measured during swallowing before and after a fatiguing task (FT). The FT involved isometric tongue pressure generation to maintain a load of 60% of the maximum pressure. Muscle synergy analysis was used as an event detector, and the 44-channel sEMG signals were divided into two analysis ranges: one containing muscle activity mainly during the oral phase (Pre-indexA) and the other containing muscle activity mainly during the swallow reflex (Post-indexA). The sEMG signals of each range were converted to swallowing pattern images, and the similarity of the images before and after FT was calculated as the Euclidean distance (ED), using a convolutional neural network and kernel principal component analysis. The results revealed that the normalized ED in Pre-indexA differed significantly between the younger and older adults. This finding suggests the possibility of quantitatively evaluating differences in age-related muscle fatigue from sEMG signals recorded during swallowing.