Sparse OFDM Channel Estimation Based on Regular Pilot Grids

Several contributions have been made on using Compressed Sensing (CS) for sparse channel estimation in wireless communications. Therein the focus is mainly set on designing transmission systems towards the application of CS channel estimation, which includes for example the proposal for random pilot patterns in OFDM systems. However, most state of the art OFDM transmission systems are using regular pilot patterns, which is beneficial for classical MMSE channel estimators. This constrains a CS-based channel estimator in its ability to choose frequency measurements at random. With the aim of using CS-based OFDM channel estimation for existing standards we examine the effect of these constraints, and we give some necessary conditions for the existence of a unique solution and good reconstruction properties. In particular, we study the effects at the example of the DVB-T standard, where we will see that it is possible to obtain reconstruction performance comparable to using random pilot patterns.