Bowel sounds is an important physiological parameter of distinguishing the gastrointestinal motility dysfunction. Auscultation of bowel sounds provides a noninvasive way for clinical diagnosis, but it is also easily affected by environmental noise. In this study, a novel higher-order-statistics (HOS)-based fractal dimension algorithm was proposed for detecting noisy bowel sounds. By using the nature of preserving non-Gaussianity for higher order statistics technique, the proposed method can effectively detect bowel sounds under different noise conditions, and its performance is insensitive to the change of noise type and noise level.