<p>A real-time diagnostic framework is developed for detecting early-stage stator inter-turn short-circuit (ITSC) faults in inverter-fed three-phase induction motors. The approach employs multiresolution Discrete Wavelet Transform (DWT) to extract time–frequency features from the stator current signals and uses a Support Vector Machine (SVM) classifier to determine fault severity. Compared to traditional Fast Fourier Transform (FFT)-based Motor Current Signature Analysis, DWT effectively captures transient and non-stationary characteristics of incipient faults. The Euclidean (L₂) norm of the wavelet coefficients is used within a Gaussian (RBF) SVM kernel to enhance classification generalization and create smooth decision boundaries. Unlike previous offline studies, the proposed system is implemented on a low-cost STM32 microcontroller as a hardware-in-the-loop prototype, utilizing optimized sixth-level db1 wavelet coefficients. This allows fully real-time fault detection without the need for additional data acquisition hardware. The framework is experimentally validated under different inverter switching frequencies, load conditions, and noise environments. The results show high diagnostic accuracy (~98–99%), low detection latency (&lt;20 ms), and strong resilience to additive noise, as evidenced by confusion-matrix analysis and latency measurements. The main contribution of this work is the integration of multi-level DWT feature extraction with SVM-based classification in a microcontroller-based real-time setup, accounting for computational and memory limitations. In contrast to existing embedded solutions, which often rely on simplified features, offline processing, or lack quantitative performance evaluation, this framework provides real-time metrics, noise robustness assessment, and latency analysis, establishing a cost-effective and reliable solution for real-time motor fault diagnosis.</p>