Papadimitriou S, Papadopoulos V, Gatzounas D, Tzigounis V, Bezerianos A
Dept. of Medical Physics, School of Medicine, University of Patras, Greece.
Stud Health Technol Inform. 1997;43 Pt B:561-5.
The present paper deals with the performance and the reliability of a Wavelet Denoising method for Doppler ultrasound Fetal Heart Rate (FHR) recordings. It displays strong evidence that the denoising process extracts the actual noise components. The analysis is approached with three methods. First, the power spectrum of the denoised FHR displays more clearly an 1/fa scaling law, i.e. the characteristic of fractal time series. Second, the rescaled scale analysis technique reveals a Hurst exponent at the range of 0.7-0.8 that corresponds to a long memory persistent process. Moreover, the variance of the Hurst exponent across time scales is smaller at the denoised signal. Third, a chaotic attractor reconstructed with the embedding dimension technique becomes evident at the denoised signals, while it is completely obscured at the unfiltered ones.
本文探讨了一种用于多普勒超声胎儿心率（FHR）记录的小波去噪方法的性能和可靠性。结果有力地表明，去噪过程能够提取实际的噪声成分。分析采用了三种方法。首先，去噪后的FHR功率谱更清晰地显示出1/fa标度律，即分形时间序列的特征。其次，重标度分析技术揭示了在0.7 - 0.8范围内的赫斯特指数，这对应于一个具有长记忆的持续过程。此外，去噪信号在不同时间尺度上赫斯特指数的方差更小。第三，用嵌入维数技术重构的混沌吸引子在去噪信号中变得明显，而在未滤波的信号中则完全模糊。
