Robust low-complexity blind frequency-dependent I/Q estimation and compensation

A low-complexity blind frequency-dependent I/Q imbalance compensation scheme is proposed based on linear prediction (LP) method. It provides non-iterative processing which reduces computational complexity. No special training sequence is used and I/Q mismatch calibration is performed using any received sequence. Frequency-dependent (FD) I/Q imbalance can be compensated simply by matching the filter responses of the I and Q branches. This is achieved by estimating the inverse impulse responses of the I and Q branches using LP and then finding the `difference´ filter that can be used in one of the branches to match the other branch. Using simulations, we show that the proposed method has at least 5-7 dB better image rejection ratio (IRR) performance compared to other blind FD I/Q estimation techniques in the literature, for a realistic mismatch scenarios. Moreover, the computational complexity of the proposed technique is shown to be 20x less than the least complex blind estimation scheme in the literature for an 8-tap compensation filter. Finally, we propose the use of a noise-generator for offline-calibration.