Three-phase induction motors are widely employed in industrial production processes, due to their low costs of acquisition and maintenance as well as their adaptation in front of different coupled load conditions and robustness for use in harsh environments. Methods that detect an incipient fault when these motors are in operation, even before significant faults may occur, have received special attention from the industry. Thus, a proper diagnosis can lead to a reduction of process losses, service time out, and equipment replacement, also ensure consistent and reliable operation of industrial systems. Thus, this article presents a comprehensive study concerning the detection of bearing faults in induction motors using information theoretical measurements and intelligent tools. Various motor operation conditions are considered, such as variations in coupled load torque and two kinds of power supply: line connected or inverter fed. This article proposes the extraction of fault characteristics in time domain, based on mutual information measurements between two-phase current signals. The method is evaluated under various fault severity levels and the performance of three different pattern recognition techniques were compared: artificial neural networks, specifically multilayer perceptron, k-nearest neighbors, and support vector machines. The experimental results presented in this article validate the robustness and effectiveness of the proposed methodology.