Feature extraction of scraping noise of the car seat electric track based on sparse representation

Abstract

Abstract With the development of the science and technology, the demanding of convenience and comfort when driving a car are getting higher and higher. As an important part that has a great influence on the comfort of the car driving, the car seat is also developing towards the electric direction. The electric seat is becoming the choice of more and more manufacture of automobile. However, the product quality inspection of the electric track of the car seat is still in a relatively preliminary stage. At present, the detection of the unqualified track product is mainly performed by workers listening in an anechoic room. This will lead to the high cost of construction of the anechoic chamber, low quality inspection efficiency, and instability by different quality inspection personnel. In this paper, we focus on the scraping noise detection of the electric track product, as scraping noise is the main cause of the product disqualification. Since scraping noise often features a higher energy in the high frequency band, we adopt the Short Time Fourier Transform (STFT) to acquire the total energy at different time duration. Then we model the acquired time series as summation of one sparse signal, one low-passing signal, and one noise signal. Based on the property of the three parts, the objective function is established and solved with Majorization-Minimization (MM) approach. The effectiveness of the proposed scraping detection method is verified with the real track vibration signal, and the superiority of the proposed method is verified via the comparison with other impulse detection method.