【摘要】 基于信号奇异谱分析技术 ,提出了奇异熵的概念 ,分析了信号信噪比与奇异熵间的内在联系 ,并提出了一种根据奇异熵增量渐近特性来对信号奇异谱降噪阶次进行确定的有效方法。仿真结果表明 ,该方法可以使信号奇异谱降噪阶次得以合理确定 ,在极大程度上确保了降噪后的信号所含信息的完整性和获得最大信噪比改善 ,这对信号特征的准确提取和可靠分析是至关重要的更多还原

【Abstract】 Lü et al proposed in 1999 a new noise reduction technique based on singularity spectrum . In our opinion, it is better than previous noise reduction techniques in many respects. On the other hand, there is still room for improving on this new technique. Ref. did not explain how to determine the order of singularity spectrum to be used. If order selected is too high, full reduction of noise is not achieved. If order selected is too low, the completeness of signal suffers. Fig.3 shows the effect of order of spectrum on the treatment of a specific stochastic noise containing signal: when order is taken to be 5, the signal is somewhat distorted; when order is taken to be 20, noise reduction is unsatisfactory; when order is taken to be 6, the signal is not distorted and noise reduction is good. The key to our improvement is that we propose the concept of singularity entropy for determining the order of singularity spectrum. Singularity entropy, as is used by us, reflects the totality of information in a noise containing signal. Through investigating the variation of asymptotical tendency of increment of singularity entropy with the order of singularity spectrum, we developed a method for determining the order of singularity spectrum to be used. Simulation results, given in Fig.5, show that our improvement of the singularity spectrum method with singularity entropy can even maintain the completeness of a noise containing Doppler signal while effectively suppressing its noise.更多还原