Adaptive equalization based on watermarking

Abstract

This paper presents an adaptive equalization method based on the use of a watermark as reference signal. Instead of interrupting communication periodically to transmit training sequences, a watermark is transmitted continuously through the channel along with the user signal. By analyzing the received watermark, the equalization filter is continuously adapted to match the channel characteristics. When dealing with correlated user signals, a prediction-error filter is introduced to whiten the equalizer output before comparing it to the reference signal (the watermark), allowing for the appropriate inversion of the channel transfer function. Simulation results are presented to demonstrate the viability of this approach.