基于小波—循环谱的新型识别技术

A New-Style Recognition Technology Based on Wavelet and Cyclic Spectrum

在通信系统中经常遇到的许多非平稳信号都具有循环平稳特性,可以用循环谱的方法来识别。但当信号本身的频谱结构比较复杂、噪声成分过大或噪声本身也具有循环平稳特性时,解调效果就不是很理想。针对循环谱识别信号的这一局限性,文中在介绍循环谱原理和识别特性的基础上,研究了几种典型通信信号的谱相关性,并提出了基于小波包与循环谱的识别技术,即先采用小波包进行软阈值消噪,再结合各自不同的循环谱特性进行识别。通过大量实验仿真表明,该方法具有很好的噪声抑制能力,能达到较好的识别效果。

In some communication systems, there are many non-stationary signals which are cyclostationary. They can be distinguished with different cyclic spectrums. However, when the signal spectrum is complex comparatively, or the noise is excessive or it also has the cyclostationarity characteristic, the demodulation results are not satisfactory. To be aimed at this limitation, this paper introduced the theory and recognition characteristic of cyclic spectrum and studied the spectral correlation of some typical communication signals. Then it proposed the recognition method based on wavelet package and cyclic spectrum analysis, which uses soft thresholds to denoise with wavelet package first, and recognizes them with different cyclic spectrums. The experiment results showed that this method has a better capability of noise rejection and it can recognize signals effectively.