A method is presented to determine the acoustic masking effects of man-made noise on marine mammal communication. In particular, the interference of icebreaker related noise with beluga whale vocalizations is studied. Captive beluga whales have been trained for acoustic experiments during which they try to detect beluga vocalizations in various noisy backgrounds. In a stop/go manner, the animals indicate whether or not they can discriminate call from noise. Results are that bubbler system noise, generated when an icebreaker ejects high pressure air into the sea in order to push ice debris away, has the worst masking effects followed by propeller cavitation noise, generated when an icebreaker is stopped by an iceridge. Naturally occurring thermal icecracking noise has the least masking impact. Based on the experimentally collected data, computer software was developed to model the whale's auditory abilities. Means of adaptive noise cancellation greatly outperform the whale and are hence not considered to take place in the whale's brain. A backpropagation neural network showed the best similarity to the whale's performance and classified the noises in the same order of disturbance as the whale.