Although the successful use of various variants of the Discrete Wavelet Transform (DWT) in the field of fault detection in Direct On-Line (DOL) fed Induction Motors, new challenges rise with the widespread use of Variable Frequency Drive (VFD) in industrial application. Regardless of the fault detection technique, the permanent variation in the fundamental frequency and therefore the fault frequencies of the analyzed signal imply a permanent re-tuning of DWT in terms of coefficient bandwidth and mother wavelet parameter. This tuning is a trial-and-error process which rises practical problem in the context of on-line Condition Monitoring Systems (CMS). Considering these limitations, the DWT use is non-tailored to the context of a VFD fed IM on-line CMS. In this paper, a novel auto-tunable DWT called the Windowed Harmonic Wavelet Packet Transform (WHWPT) using drive's data is presented in order to overcome the aforementioned drawbacks. The new approach has been validated experimentally using broken rotor bar fault in a VFD fed 3-kW Double Cage IM under a power supply at different fundamental frequencies. Fault feature extraction results show the robustness and the flexibility of the proposed approach through clear distinction between healthy and faulty motor conditions even for several fundamental frequencies.