Separation, Allocation and Psychoacoustic Evaluation of Vehicle Interior Noise 2019-01-1518

Besides optic and haptic criteria the interior sound highly influences the quality impression. Especially the occurrence of separately audible noise components is commonly associated with an insufficient product quality. As a result, the reduction of disturbing noise components is a key factor for the overall product quality. Since the acoustic optimization is complex and time consuming process, the need for an analysis tool which identifies automatically disturbing engine noise components within the vehicle interior noise is high. For this reason, a novel analysis tool is developed which extracts tonal and impulsive engine noise components from the overall engine noise and evaluates the annoyance of each noticeable engine component automatically. In addition, each disturbing noise is allocated to the emitting engine component. It is then possible to listen to each engine component noise individually and synthesize a target noise by superimposing manually weighted component noises. The noise separation into noise fragments is performed by means of the non-negative matrix factorization and image processing tools. These are then clustered according to their time correlation or other similarity-determining parameters. Classification algorithms are then trained using chosen features to provide a noise source allocation. The extracted noise components as well as sound mixtures of those components are assessed by a psychoacoustic pleasantness rating model. This model is based on a multiple regression analysis between experimentally acquired pleasant values and objectively calculated psychoacoustic parameters as, e.g., loudness, sharpness, tonality. It also considers auditory masking. Components whose sound pressure levels are below their masked thresholds will not contribute to the overall pleasantness which means that they can be neglected in further analyses.