Segmentation is an important stage in signal analysis, and its performance evaluation plays a significant role in the efficiency of the subsequent steps, such as extraction of descriptive features and classification. There are a large number of approaches to segment signals. The performance of each of them remarkably varies when the signal changes. In this present study, two novel algorithms, which use the probability and fuzzy concepts, are proposed to combine several well-known existing signal segmentation approaches. The simulation results confirm the efficiency of the proposed approaches using the synthetic and real electroencephalogram signals.