Detection of transients using the nonstationary bispectrum

Abstract

An algorithm based on the nonstationary bispectrum is proposed for detection of sampled finite duration signals. The detection algorithm uses a combination of coherent and incoherent smoothing in the frequency domain to produce a test statistic which is suitable for display as a spectrogram. A model transient waveform imbedded in Gaussian noise is used to evaluate the detection performance of the proposed detector. Detectors based on the spectral correlation and power spectrum are used as metrics to evaluate the relative effectiveness of the proposed detector. Performance is investigated both as a function of signal-to-noise ratio for a fixed transient duration and processing length and as a function of transient duration/processing length mismatch at a fixed signal-to-noise ratio. The benefits and limitations of the proposed test statistic based on the nonstationary bispectrum relative to the other detectors investigated will be noted. [Work supported under contract with Space and Naval Warfare Systems Command.]