Performance of underwater source separation using narrowband and broadband signal at mini underwater test tank with varied in salinity and temperature

SONAR is one of underwater communication by using acoustic wave which having smaller attenuation value in underwater than others wave. However, SONAR received mixing information from the source caused by noise from ship, offshore activity and seismic noise. Therefore, in this paper we propose a method using Independent Component Analysis (ICA) to separate the mixing signal to get the desired signal. Sound source separation is performed at mini underwater test tank in Vibration and Acoustics Laboratory using the same number of source and sensor configuration. The variables are varied based on salinity, temperature and sensor displacement changes. The separation process was carried out using TDICA, FDICA, and MSICA methods. From the mixed sound recording results obtained the best separation results using FDICA and MSICA methods for each variable change. The average MSE score was 0.0890 and 0.0899 for the FDICA and MSICA methods, while for the TDICA method the average score of MSE 0.1692 was obtained. Along with the increase in distance and changes in propagation medium, the quality of the separation is decreased.