Pulse Detection In Acousto-Optic Channelized Receivers

The large time-bandwidth products of acousto-optic spectrum analyzers require dedicated post-processing hardware for real-time signal parameter estimation with low false alarm rates. Parameters of interest include center frequency, time-of-arrival, time-of-departure, and amplitude. Maximum likelihood parameter estimation can be performed with a spectrogram correlation process, which is a 2-d matched filter operating on the time-frequency history of the spectrum analyzer. The filter is matched to the known covariance of the signal. A data collection board was constructed to allow capture of 256 channels at a readout rate of up to 15 μS. This data was used to examine sub-optimal implementations of the 2-d matched filter. A high-accuracy 2-d peak detection algorithm was developed to assess the filter performance. Experimental results are presented, and design tradeoffs are shown to exist between filter complexity and accuracy of parameter estimation.