The condition monitoring of bearings has gained great importance in recent years to increase reliability and reduce production loss. Many monitoring techniques have been proposed based on different intelligent techniques and feature extraction schemes. In this study, a combined decision algorithm has been developed based on feature set that composed of statistical variables and linear prediction coefficients of time domain vibration signals. Artificial intelligent techniques, namely artificial neural networks, adaptive neuro-fuzzy inference systems and support vector machine were employed together to develop a decision making algorithm that classify the type and severity of bearing faults. Although each method can be used alone for data classification in the developed models with a limited performance, the proposed decision algorithm combines decision of each method with a synergy according to the majority of the decisions. Based on the experimental results, the proposed scheme outperformed the three methods when used alone.