Fault diagnosis in electrical drives can be very difficult, especially when the input data comes from an operating electrical machine. The occurrence of Inter-Turns Short-Circuit (ITSC) fault is one of the most dangerous electrical machine failures, and if these are not detected at an early stage of development, they can result in serious consequences, both in terms of repair cost and safety. In this context, an effective on-line approach for diagnosisng ITSC fault is proposed, which is based on the computation and monitoring of a specific severity factor, defined as the ratio of positive and zero symmetrical components. This approach is implemented in a LabVIEW environment and employs the Short Time Least Square Prony's (STLSP) method. This does not require the determination of motor parameters and just involves the use of voltage sensors. Several tests were performed on a Synchronous Reluctance Machine (SynRM), for various operating conditions (healthy and faulty). The obtained results confirm the effectiveness of the proposed technique for diagnosing ITSC faults, with high reliability, rapidity, and accuracy.