Road & wind noise contribution separation using only interior noise having multiple sound sources -Accuracy improvement and permutation solution-

Abstract

A contribution separation technique for estimating road noise and wind noise contributions to vehicle interior noise without using any sound source information was considered when each noise was from multiple sources. In the test, road, wind or mixed noise sources were reproduced from four speakers placed at different positions. The mixed interior noises were recorded at two separate positions surrounded by the speakers. Independent component analysis (ICA) was then applied as the contribution separation method for obtaining them at the target position (driver seat position). As the original condition, ICA was firstly applied to the recorded noises at the driver seat (target noise) and at the assistant seat (reference noise). The result revealed that the estimated contribution was inaccurate due to the low accuracy at the low frequency and the permutation problem, in which the relationship between the calculated contribution and the actual contributions is not retained with changes in frequency. Next, the reference signal was changed to the other noise having lower correlation to the target for improvement of the accuracy and the permutation solution was considered by using the time trend correlation of the calculated contributions of dominant independent component. The modified ICA technique with the permutation solution could estimate more accurate contributions than the original condition.