为了提高振动信号特征参数提取的准确性，本文结合离散小波变换与梅尔频率倒谱系数提出一种特征参数提取方法。先对采集的振动信号进行预处理，对预处理后的信号进行离散小波变换，得到小波系数。对小波系数进行快速傅里叶变换，求出其频谱，得到各层小波系数的样点数。按照频率的高低顺序，将各层小波系数合成完整频谱，计算系数能量，求出mel对数功率谱。然后对mel对数功率谱进行离散小波变换，提取mel离散小波变换谱振幅作为振动信号特征参数。实验表明，该方法提高了振动信号特征参数提取的准确性。%In order to improve feature extraction accuracy of the vibration signal, this paper presents a characteristic parameter extraction method that combines the discrete wavelet transform and Mel Frequency Cepstral. Before the vibration signal acquisition preprocess, preprocessed signal discrete wavelet transform, and acquie the wavelet coefficients. Of the fast Fourier transform for the wavelet coefficients, the spectrum is obtained, the layers of the wavelet coefficients to obtain the number of samples. In descending order of frequency, the wavelet coefficients of the synthesis of the complete spectrum of each layer, calculation of the energy coefficients, and mel logarithmic power spectrum is obtained. Then mel logarithmic power spectrum of the discrete wavelet transform, discrete the wavelet transform to extract mel spectral amplitudes as a vibration signal characteristic parameters. Experiments results show that the proposed method improves the characteristic parameters extraction accuracy of the vibration signals.
展开▼
