Frequency Estimation with Co-prime DFT Filter Bank and Half Nyquist Frequency Shift

Abstract When processing wide sense stationary (WSS) signals, the sampled data rate of conventional MN bands DFT filter bank is at least MN, while co-prime (co-prime DFTFB) only needs M+N, that sharply lowers sampling rate. Here, M and N are pair integers. However, practical FIR imperfections correlation detection mode arouse spurious peaks in signal power spectrogram, greatly limits application DFTFB. In this paper, we propose a modified DFTFB by adding real valued multiplier to original one, where both odd numbers. Together with DFTFB, get 2MN bank. simulation section, two BPSK signals random carrier frequencies employed. The results probability versus SNR curves through 500 Monte Carlo experiments verify effectiveness our method.