摘要
在对生物医学信号时间序列进行复杂度分析时,粗粒化预处理有可能会造成丢失原始信号中所蕴含的信息,甚至在某些情况下根本改变原信号的动力学性质。用计算机计算时的量化过程也是一种粗粒化,因此也有这类问题。通过对近似熵和我们所定义的C0复杂度这两种复杂度在不同量化精度下对一些典型时间序列复杂度分析的比较研究,发现一般说来量化精度对复杂度分析的影响不是很大,仅当对原始信号进行二值化等极端情况下,才会显著改变原信号的复杂性。对脑电信号进行计算表明上述结论是实际可取的。
Our previous paper indicated that a coarse graining preprocessing might result in a loss of information in the original signal, when a complexity analysis was done for biomedical signals; and even in some cases the dynamic property of the original signal might be changed radically due to such preprocessing. The quantization process is also coarse graining processing, thus the risk mentioned above should be considered when the original data were quantized. Compasing the influence of the degree of quantization on C0 complexity and on approximate entropy (ApEn) values calculated for some typical time series, we found that, generally speaking, such influence is not significant, except for some extreme cases such as the binary quantization. Practical calculation of EEG data confirmed the above conclusion.
出处
《生物物理学报》
CAS
CSCD
北大核心
2000年第4期707-710,共4页
Acta Biophysica Sinica
基金
国家自然科学基金
北京认知科学开放实验室资助
关键词
脑电
量化
复杂度分析
EEG
Quantization
Complexity analysis
