Abstract: A noise detection signal detected at an evaluation point is filtered by a low pass filter, so that an evaluation waveform signal is extracted. A real signal mother wavelet of complex type is derived from the extracted evaluation waveform signal. Using this mother wavelet, an abnormal waveform signal stored in advance is processed by wavelet transformation. Then, a correlation value is calculated between the abnormal waveform signal and the mother wavelet. The correlation value is compared with a predetermined criterion, so that the presence or absence of abnormality in a sound source or vibration source corresponding to each abnormal waveform signal is determined.

Abstract: A detection signal of sound or vibration detected at an evaluation point is filtered so that an evaluation waveform signal is extracted. Then, a mother wavelet is derived using this. On the other hand, each of detection signals of sound or vibration detected at a plurality of candidate points is filtered so that each candidate waveform signal is extracted. These candidate waveform signals are processed by wavelet transformation using the derived mother wavelet. Then, on the basis of the result, instantaneous correlation is calculated between each candidate waveform signal and the mother wavelet. Then, on the basis of mutual comparison of the calculated instantaneous correlation, a generation source of the sound or vibration that appears at the evaluation point is identified.