As is well-known, it is very often that an objective noise fluctuation is contaminated by the other noise sources. That is, in the actual acoustic measurement, the acoustic signal is measured by a sound level meter under the mixture of several sound sources. From the above point of view, we pay attention to blind source separation or extraction of independent source signals from their mixtures. Blind source separation is solved based on Independent Component Analysis (ICA) by utilizing mutual independence among source signals. In this report, by introducing ICA, we propose a probabilistic estimation method for the objective sound level fluctuation from the mixture of several sound sources.