Creep Groan Noise Classification

Anže Železnik, Severin Huemer-Kals, Jure Murovec, Jurij Prezelj

Research output: Chapter in Book/Report/Conference proceedingConference paper


Creep groan is an unpleasant noise caused by stick-slip during braking at low speed and low to medium brake pressure. To automatically identify creep groan phenomena with a machine learning algorithm, the number of features extracted from accelerometers on the brakes needs to be as small as possible. This paper focuses on a brute force method for selecting the optimal combination of psychoacoustic features that can be used with the self-organising map to identify the creep groan noise of a vehicle's brakes. The number of classes was selected using principal component analysis and plotting the distances between two randomly selected data points. The results were evaluated using the maximum distance between SOM neurons and the R2 coefficient between the classified data and the subjective ratings of the brake noise. The results of this study show that the best psychoacoustic features for identifying creep groan are sharpness and fluctuation strength, and that unsupervised classification is more reliable than subjective classification of braking noise.
Original languageEnglish
Title of host publicationThe 9th Congress of the Alps Adria Acoustics Association – Conference Proceedings
EditorsMesterházy Beáta, Mikló Márkus
Place of PublicationBudapest, Hungary
Publication statusPublished - 24 Sept 2021
Event9th Congress of the Alps Adria Acoustics Association: AAA 2021 - Budapest, Hungary
Duration: 23 Sept 202124 Sept 2021


Conference9th Congress of the Alps Adria Acoustics Association
Abbreviated titleAAA 2021

Fields of Expertise

  • Mobility & Production


Dive into the research topics of 'Creep Groan Noise Classification'. Together they form a unique fingerprint.

Cite this