Bearing failures can have a significant impact on machineries. Proper analysis is essential to ensure smooth operation and timely error detection. This research aims to detect defects in rolling element bearings (REBs) using the SPWVD-CNN (Smoothed pseudo-Wigner-Ville distribution—Convolutional Neural Network) approach, which combines machine learning, signal processing and wavelet transform. The main objective of this study is to develop a condition monitoring approach for machine components, predicting the condition of REBs using advanced analysis techniques such as deep learning. By using machine learning algorithms, potential defects can be identified early, allowing corrective actions to be taken and reducing the risk of costly outages and catastrophic failures/accidents. The research methodology involves collecting and converting data into the required format. The Case Western Reserve University bearing data have been used to extract statistical features. The SPWVD technique is applied to extract features generating spectrograms, which are then characterized and labeled. This tagged dataset (i.e., supervised learning) serves as input to a Convolutional Neural Network (CNN) model. The accuracy of the CNN model is evaluated to determine the occurrence of defects in the REBs. The Experimental results show that the proposed method achieves an accuracy of 100% using the SPWVD. Furthermore, the proposed SPWVD-CNN model is compared to existing machine learning models to analyze deviations in accuracy. The results demonstrate the effectiveness of the SPWVD-CNN approach in accurately detecting bearing defects and highlight its potential as a reliable condition monitoring technique.