This work shows the importance of parameterizing and extracting the characteristics when describing vibration signals. For this purpose, we use techniques based on calculating characteristics in the time and frequency domain and genetic algorithms to broadly explore the solution space to guarantee convergence to an optimal value. This research aims to determine better performance characteristics when analyzing a vibration signal on bearings to predict failures. This paper describes a methodology to analyze signals using different representation spaces such as time, frequency, and time-frequency, based on which we compare various characteristics to find ones with a higher impact on signal analysis. Using a fitness function to determine the analysis reliability considering each iteration efficacy confirms the new method's effectiveness, and then a new group of characteristics will be considered and found using genetic algorithms. The research results improve the methodology to analyze signal problems and can apply as a tool to study future failure diagnosis problems. Keywords: Genetic Algorithms, Databases, Fusion of Characteristics, Bearings DOI： https://doi.org/10.35741/issn.0258-2724.58.1.68