Inter-turn faults are one of the most important issues in line-start permanent magnet synchronous motors (LSPMSMs). To date, steady-state operating conditions of LSPMSMs have been analyzed in studies focusing on diagnosing this fault. However, no studies have been conducted for nonstationary operating conditions, such as variable speed or variable load conditions, which are common in the industry. This paper presents a novel approach for fault harmonic component (FHC) tracking based on transient-MCSA to diagnose inter-turn faults in an LSPMSM operating under nonstationary conditions. First, the inter-turn failure model of LSPMS was analyzed to show the fault effect on motor current, and the most dominant inter-turn FHC in the order and frequency domains were determined. Then, using Gabor-OT, interturn FHC signals were extracted from motor current signals in the frequency domain and reconstructed in the time domain. Intrinsic mode functions (IMFs) were calculated by decomposing the reconstructed FHC signals (RFHCSs) via ensemble empirical mode decomposition. Using the energies of RFHCSs, calculated based on Gabor coefficients, and the Kullback-Leibler divergences of the selected IMFs, diagnosis of the inter-turn faults and detection of the severity of the faults was performed. The results are in agreement with the results in the literature, thus showing that the proposed method is a successful and useful means for detecting interturn faults in LSPMSMs.