Robust DFT-based generalised likelihood ratio test for underwater tone detection

The discrete Fourier transform (DFT) of observation sequence is usually used in the detection of tonal signal. A tone without any prior information cannot be guaranteed to lie exactly on the discrete DFT frequency bin, which will degrade the performance of the tonal detector based on DFT. In this work, the authors study the detection problem for a tone deviated from the discrete frequency bin. A generalised likelihood ratio detector utilising the information of adjacent frequency bins is designed. Theoretical analysis, computer simulations, and lake experiment show that the proposed detector is robust against the deviation and is more suitable for the tone detection in practical circumstances.