Knocking recognition in engine vibration signal using the wavelet transform

The authors propose pattern recognition using wavelet parameters for detecting a physical signal, i.e., knocking, in engine vibration. It is shown that wavelet transform parameters of the vibration signal improve knocking detection, especially at high speed, for slow signal-to-noise ratio (SNR). The knocking signal is characterized by the sum of several resonance frequencies. This method should give good results for other signals which have the same characteristics.<<ETX>>