Fault identification in engineering materials is strongly impacted by the growing usage of ultrasonic-based detection techniques. The fundamental goal of ultrasonic inspection of engineering materials is to quickly and correctly discover and characterize internal flaws and imperfections. Non-destructive testing (NDT) is complex and time-consuming, thus the tester’s experience and understanding are essential. After some effort, humans’ eyes and brains can classify a broad variety of complicated patterns. However, factors such as fatigue and lack of concentration can have a significant influence on performance. Traditional NDT approaches based on heuristic experience-based pattern recognition methods have considerable shortcomings in terms of cost, length, and unpredictable analysis, resulting in inconsistent outcomes. To characterize the flaws, signals from a pulse receiver are analyzed. EMD, DWT, and Stockwell Transform are cited in signal processing literatures because these signals are non-stationary in nature. The Hilbert–Huang transform is castoff to characterize flaws in this paper, and the PSD is treated to characterize the features. The average power of volumetric defects is found to be higher than that of planar defects.