Novel Multiwavelet-based LPC Random Forest Classifier for Bluetooth RF-Fingerprint Identification

An innovative bluetooth radio-frequency (RF) fingerprint identification scheme using the random forest classifier involving multiwavelet-based linear-predictive-coding (LPC) features is introduced in this paper. In our proposed approach, finite-element multiwavelet with an arbitrary multiplicity (MWAM) is first constructed to decompose an RF signal emitted by an electronic equipment into multiple subbands. Next, LPC coefficients, which can be employed to mitigate the background noise, are further estimated from these subband signal sequences. Such multiwavelet-based LPC coefficients will be utilized as the features of the adopted random-forest classifier to recognize the bluetooth RF fingerprints emitted from different wireless transmitters. Monte Carlo simulation results demonstrate the effectiveness of our proposed new RF fingerprint identification technique.