Auditory-inspired Interval Statistic Receivers for Passive Sonar Signal Detection

The narrowband detector model used in passive sonars consists of a band-pass filter followed by a quadrature receiver, and so operates exclusively on signal power. The human ear suggests an alternative approach to narrowband detection based on signal phase, which is captured by phase locking in auditory nerve fibres. According to these principles, three "interval detectors" - detectors using information in zero crossing intervals - are developed and evaluated alongside a traditional power detector. The detectors all apply a maximum a posteriori criterion and assume Gaussian statistics. The power detector is shown to outperform the interval detector when the signal is centred on the analysis filter. When the signal frequency is displaced from the centre, interval detection improves and power detection worsens. Furthermore, a critical displacement exists, beyond which interval detectors are superior, and become useful for discriminating a weak signal in the filter tail.