Mean texture depth measurement with an acoustical-based apparatus using cepstral signal processing and support vector machine

In this study, an acoustical-based macrotexture measurement method that uses the friction-induced mechanism of tire/road noise has been proposed. This mechanism influences the frequencies above 1 KHz and is the dominant mechanism for frequencies above 3 KHz. The airflow due to this mechanism passes through the macrotexture which can be represented using convolution in the time-domain of the airflow, which is viewed as the input signal, and macrotexture, which is considered as the transmission channel. Using the Cepstral processing method, the related features to macrotexture airflow can be extracted and the macrotexture can be evaluated. To investigate the proposed method, interaction noise was measured for six uniform non-porous asphaltic test tracks with different macrotextures. The tire/road noise signal is passed through a bandpass filter (3–5 KHz). By liftering the cepstrum components in the range of 2–50, the feature vector has been obtained and then used as input to a multiclass SVM, resulting in a precision error of 4%.