Abstract
In this study, we applied an independent component analysis (ICA) to separate the contributions of vehicle interior noise by using only response signals. To verify the applicability of ICA, we utilized engine and wind noise and integrated them into a simple simulation with three conditions (synchronous, asynchronous, and asynchronous with frequency characteristic), and identified their separate contributions by applying ICA to these mixed response signals. Results show contributions were calculated accurately while using time-domain ICA for the case in which source signals were synchronously mixed, and it was found that accuracy was low when the signals were asynchronously mixed. Frequency-domain ICA was found to separate the contributions correctly. When the source signals were asynchronously mixed with specific frequency characteristic, permutation problems occurred, in which the matching between calculated independent component by frequency-domain ICA and the actual sound source changed among frequency. Therefore, a permutation solution method using a non-Gaussian characteristic of the calculated contributions was proposed. As a result, the contributions could be calculated accurately.