Pilot Pouring in Superimposed Training for Channel Estimation in CB-FMT

Cyclic block filtered multi-tone (CB-FMT) is a waveform that can be efficiently synthesized through a filter-bank in the frequency domain. Although the main principles have been already established, channel estimation has not been addressed yet. This is because of assuming that the existing techniques based on pilot symbol assisted modulation (PSAM), implemented in OFDM-like schemes, can be reused. However, PSAM leads to an undesirable loss of data-rate. In this paper, an alternative method inspired by the superimposed training (ST) concept, namely pilot pouring ST (PPST), is proposed. In PPST, pilots are superimposed over data taking advantage of the particular spectral characteristics of CB-FMT. Exploiting the sub-channel spectrum, the pilot symbols are poured in those resources unused for data transmission. This spectral shaping of pilots is also exploited at the receiver to carry out channel estimation, by enhancing those channel estimates that exhibit a low data interference contribution. Furthermore, a frequency domain resource mapping strategy for the data and poured pilot symbols is proposed to enable an accurate estimation in strongly frequency-selective channels. The parameters of the proposed scheme are optimized to minimize the channel estimation mean squared error (MSE). Finally, several numerical results illustrate the performance advantages of the proposed technique as compared to other alternatives.