Self-noise cancellation methods applied to acoustic underwater sensors

Self-noise generating mechanisms can be categorised in noise produced by the equipment and the platform. Equipment noise has characteristics of being instantaneous in nature, e.g. transients such as tube noise. The noise produced by the sensor platform includes noise from the machinery and hydrodynamic noise generated by the motion of the platform in the water. A signal processing perspective on reducing the negative impact of self-noise is taken. Two promising self-noise cancellation methods are investigated. Both methods consist of a noise estimation followed by filtering. The first method estimates the noise by fitting an AutoRegressive (AR) model, while the second approach uses an Independent Component Analysis (ICA) method. A Finite Impulse Response (FIR) Wiener filter and three different adaptive FIR filters, LMS, NLMS and RLS are then applied. The methods are applied to Towed Array Sonar (TAS) signals taken from a field test carried out in the archipelago of Stockholm. The AR approach performs well regardless of which filter used and suppresses the noise by 30 dB/Hz even when the correlation coefficient between the interfering and measured noise is as low as 0.2. The ICA method reduces the noise in the main frequency band by 10 dB/Hz for all filters