In response to the issue of user dissatisfaction and fatigue caused by repeatedly receiving similar content, this paper proposes a sequence recommendation model based on Transformer and fatigue extraction, named TransFESRec. Initially, TransFESRec extracts dynamic interest representations from users" behavior sequences using a Transformer. Then, it employs the Fast Fourier Transform (FFT) to convert the behavior sequence from the time domain to the frequency domain, thereby extracting the user"s fatigue representation. Subsequently, the model combines the fatigue representation with the interest representation and inputs them into a multilayer perceptron (MLP) to learn the nonlinear relationships between them, forming a comprehensive representation of the user"s state. Finally, this integrated representation is dot-multiplied with the embedding vectors of each candidate item to determine the user"s preference level for each item. Experimental validation shows that TransFESRec effectively reduces the recommendation of content types that lead to user fatigue and outperforms other mainstream methods across multiple evaluation metrics.