This study aims to systematize knowledge on feature extraction in signal processing and propose a new approach emphasizing the importance of association rules. The paper begins by adapting the traditional DRISP-DM model for signal processing applications, introducing the CRISP-SP model, and highlighting the critical role of feature selection in the identification process. The analysis reveals that features serve three distinct functions, leading to the categorization of feature extraction methods into three main areas: methods focused on discovering association rules between signals and the tested samples or processes, techniques aimed at accurately representing signals using a limited set of attributes, and methods designed to select the optimal set of attributes for identification. The study comprehensively reviews various feature extraction methods, examining their functionalities, strengths, and limitations. It also explores the role of association rules and proposes an algorithm specifically designed to extract these rules from signals. Additionally, it presents relationships between these methods in the form of maps, which can help researchers select the most appropriate techniques for specific scenarios. The paper focuses on general methods. Therefore, the content has been verified by specialists from diverse fields related to signal processing. The proposed signal feature extraction and selection methods enhance the decision-making process. This is important in disciplines including management and engineering, as it offers quicker and more objective analyses across various business and industry sectors.