Research of brake sound acoustic features extraction based on frequency-domain blind deconvolution

Vehicle brake will produce several noise during braking process, it will influence the life of automotive components. Meanwhile, the sharp brake noise (screaming) will seriously interfere with people´s normal life. Thus, the work of brake noise abnormal sound governance is very important. In this paper, the generation mechanism, features and influencing factors of brake are summarized. The theory and research progress on brake noise suppression and prevention are review and analyzed. For the outstanding problems of vehicle brake abnormal sound detection caused by lacking of detection and analysis methods, inconvenience of data processing and poor flexibility, a brake abnormal sound localization method based on frequency-domain blind signal processing was proposed. Its key technologies such as using unit fixed-point algorithms based on kurtosis maximization to separate and extraction complex components and solving permutation indeterminacy based on improved KL-distance were all introduced in detail. Finally, an actual brake failure acoustic signal extraction shows the effectiveness and reliability of this method.