基于希尔伯特-黄变换的白细胞信号分

白细胞信号(WBS)具有脉冲形状多样和时频特性各异的特点，目前临床上用细胞信号脉冲计数的方法来分析WBS通常难以反映其所携带的丰富的生理和病理信息，并直接影响到细胞分类问题。针对这一问题，探索能自适应地分解非线性非平稳信号的希尔伯特 黄变换算法在WBC时频分析和分类中的应用效果。通过对血细胞中的WBC进行HHT变换，获取WBC的本征模态函数（IMF）分量、分量的Hilbert边际谱以及信号Hilbert谱；利用瞬时频率、瞬时幅值等进行计算提取健康人与患者的WBS平均强度、谱质心以及能量贡献率等特征作对比分析，根据其时频特征分布规律构建用于分类实验的特征向量；采用支持向量机（SVM）分类器，对58名健康人和60名患者的白细胞实验样本进行分类实验。结果表明，该方法提取的健康人和患者的WBS分量谱质心分布、平均强度值以及能量贡献率具有较好的区分度，分类正确率到达了94.83%。HHT方法能有效提取WBS特征，可辅助临床WBS的处理和分析。

The Analysis of White Blood Cell Signal Based on Hilbert-Huang Transform

White cell signal (WBS) has various shapes of pulse and different time frequency features, which makes it difficult to extact the physiological and pathological information from WBS using cell signal pulse counting method and classify the blood cells accurately in clinics. In this work, the Hilbert Huang transform (HHT) method which can adaptively decompose non stationary and nonlinear signal was investigated to explore its application in WBS time frequency analysis and classification. Using HHT, WBC′s intrinsic mode function (IMF), Hilbert marginal spectrum of IMF and  Hilbert spectrum of WBS were obtained; Average intensity, spectral centric and the energy contribution rate of WBS of healthy people and patients were extracted and analysised through instantaneous frequency and  instantaneous amplitude. According to the distribution of time frequency features, the feature vector for classification experiments were constructed, and then support vector machine (SVM) was adopted in the classification of  WBS experimental samples of 58 healthy persons and 60 patients. Results showed that the correct ratio of WBS classification was 94.83%. In conclusion, the HHT method is effective in extracting the WBS features and may be expected to assist clinical WBS processing and analysis.