Bearings are critical components of rotating machinery and monitoring their condition is important to avoid catastrophic failures and reduce the machinery down-time. Acoustic emission (AE) is gaining ground as a complementary condition monitoring technique as it offers earlier fault detection compared with other more established techniques, such as vibration analysis or oil analysis. However, AE signals always include a significant level of noise reducing the potential of defect detection at early stage. For this reason, this paper proposes a novel envelope analysis method for bearing incipient defect detection. This method is able to identify localized defects in an incipient stage, in which the signal-to-noise ratio (SNR) is extremely low. This method combines Wavelet packet, for AE signal denoising, the Hilbert Transform (HT) for envelope extraction and autocorrelation function, to find patterns in the AE signal. An extensive experimental investigation was carried out in order to evaluate the performance of the proposed method under extremely low SNR, adding high level of noise to the signals. The results indicate that the proposed enhanced envelope method is able to detect incipient defects with 9 dB lower SNR than traditional envelope analysis.