Variable Scale Relative Entropy Detection for Non-Cooperative Underwater Acoustic Pulse Signals

This work investigates the detection of non-cooperative underwater acoustic pulse signals at low signal-to-noise ratio (SNR). A variable scale relative entropy (VSRE) pulse signal detection scheme is proposed. Different from conventional relative entropy (RE) method where an observation sequence is processed at a given scale for detection, the proposed scheme performs the processing at multiple different scales. As a result, the original RE difference vector becomes a RE difference matrix, each column corresponding to a particular scale. Before a decision is made, the VSRE difference matrix is post-processed for improved fidelity. The non-zero elements of the resulting VSRE difference matrix are divided into groups, each corresponding to a pulse signal, based on their occurrence times within the observation sequence. Within each group, the one with the maximum RE difference is chosen to determine the exact appearance time and duration of the pulse signal. The performance gain of the VSRE detector over existing RE detector has been verified by both simulation and at-sea experimental results.