A novel sound quality (SQ) evaluation method is developed to diagnose abnormal noise of interior permanent-magnet synchronous motors (IPMSM) for electric vehicle (EV). A-weighted sound pressure level (A-W SPL) of noise in IPMSM and six psychoacoustic characteristics, which include sharpness, roughness, loudness, articulation index, tonality, and fluctuation strength, are described through the objective evaluation of SQ. The paired comparison method (PCM) and the grade evaluation method (GEM) are respectively applied in subjective evaluation on relative annoyance and absolute annoyance of IPMSM for EV, and the combined application of these two methods is proposed originally. Finally, the BP neural network model for SQ evaluation of abnormal noise in IPMSM is established, the characteristics, conditions, and influencing factors of abnormal noise are effectively diagnosed. This novel evaluation method effectively solves the problem that the annoyance level is unable to be evaluated through the PCM, and the GEM relies on rich audition experience of evaluation subject. It also improves the accuracy of diagnosis of abnormal noise in IPMSM for EV.