Simulation and Detection of Intermittent Sounds in Wind Noise Tests on Automobiles

Customer complaints about wind noise in cars have led to a re-examination of the metrics used in evaluations of wind noise. A refinement to the loudness-based metric has been proposed and a validation study was conducted to determine how well the model worked in predicting the quality of wind noise in new cars. As part of this validation process a new error state was identified. Error states are the presence of noise characteristics that would make a vehicle unacceptable, and thus indicate problems that must be addressed. An example of this is whistling noise. In the validation study another sound characteristic, which is an intermittent high-frequency repetitive noise, was identified as a potential error state. In this paper, a methodology to simulate this intermittent noise is described, so that the threshold level for detection and just noticeable differences of this error state noise could be explored. The simulation method involves generating a signal of intermittent filtered white noise and adding it to a user-rated acceptable wind-noise recording. Both the intermittent noise and the original recording are filtered so that the resulting synthesized sound has a similar spectral shape to the recording with the error-state noise.