In the health monitoring of steel structures, the acoustic emission (AE) testing method is the most eminent method as it can identify the micro-level cracks in inaccessible places. AE waves are a quick release of energy inside any material that can be recorded using piezoelectric sensors. It is furthermore possible while capturing the AE signals that the sensors can record the noise signals which can result in an error. So the true information about the structure is restrained. Hence it is important to discriminate the actual AE signal from the other signals captured by the sensors. The employment of artificial intelligence will reduce human involvement and increase the accuracy of any AE testing-related applications such as damage detection or source localization in any structure. Therefore, this paper proposes a discriminating method of separating different signals captured by the sensors from the actual AE crack signals with the help of the support vector machine (SVM) algorithm. The AE crack data are collected on a 1.5 mm thick steel plate, which implements an R6∝ sensor using a pencil lead break-up test. Other noisy signals, i.e., rubbing and impact signals are generated by rubbing a small metal block and dropping it on the steel plate, respectively. The generated AE signals, sensed by the sensors are amplified since AE waves release a very small amount of energy, and then they are sent to the data acquisition system for further processing.