一种基于时频域归一化二次谱的信号特征检测方法

Detection Method of Signal Characters Based on Normalized Quadratic Spectrum in Time-Frequency Domain

针对传统的基于傅里叶变换的信号特征检测方法在频率域内没有考虑时间分辨率,而基于小波变换的方法对缓变信号特征检测困难等问题,提出了一种基于时频域归一化二次谱的信号特征检测新方法.该方法首先在频率域内对原始信号的振幅谱进行傅里叶变换,再将其归一化,获得归一化二次谱,这种二次谱的自变量具有时间量纲,它既能反映信号的频率特征,也能反映信号的时间特征.然后,基于二次谱进行特征检测,同时可采用补零或加窗等方法对数字信号进行处理,以防止频谱泄露,从而克服了单纯频谱分析的不足.针对大型杆件与石油管线传导信号的特征检测结果表明,所提方法能对被测物的完整性做出准确、有效的诊断.

The traditional signal detection method based on Fourier transform has no time resolution in the frequency domain and the detection method based on wavelet transform cannot characterize the slow-varied signal. Focusing on the problem mentioned above, a new detection method based on normalized quadratic spectrum （NQS） in the time-frequency domain is proposed. Firstly, Fourier transform of the original signal＇s amplitude spectrum is performed in the frequency domain, and then NQS is obtained by normalizing the transformation of amplitude. Afterwards, the signal detection is performed based on the obtained NQS. The supplementary zero or adding window methods, etc. can be used simultaneously to prevent from spectrum leakage. It is proved that NQS can reflect both the frequency and the time characters of the signal. This merit of NQS can overcome some pitfalls of traditional detection methods. The validity of the proposed method is verified when it is applied to the character expression of the pole and oil pipeline＇s conductive signal.