Rolling element defect identification is a difficult task. The reason being that defect on the rolling element has both rotational as well as revolutionary motion. To identify rolling element defect in a taper roller bearing, a novel signal processing scheme is proposed which results in a substantial increase in kurtosis and impulse factor of the vibration signal. The scheme constitutes a series of operations. In the beginning, the raw signal is decomposed by ensemble empirical mode decomposition (EEMD) and inverse filtering (INF). The above two stages of signal processing extract hidden impulses which are suppressed in the noise present in the experimental data. In the third stage of processing, continuous wavelet transform (CWT) using adaptive wavelet is applied to the preprocessed signal to produce a 2D map of the CWT scalogram. This transformation results in a higher coefficient in the region of impulse produced due to the defect. Finally, time marginal integration (TMI) of the CWT scalogram is carried out for defect localization. The defect frequency was evaluated with an accuracy of 97.81% and defect location was identified with an accuracy of 92%.