基于增强的谱分析和奇异值分解的T波交替检测

针对传统谱分析方法在T波交替(TWA)检测中对噪声敏感的缺点,提出将增强的谱分析方法和奇异值分解方法结合起来的TWA检测方法.该方法利用奇异值分解得到去除了噪声干扰的心电信号,克服了传统谱分析方法只能检测平稳信号且需要增大心率的缺点,强调交替水平的重要性,即增强TWA功率谱上0.5cpb处的幅值,实现对TWA的有效分析.研究结果表明：该方法对T波交替数据库中30个人工合成并含有TWA的数据的检出准确率达93.33%,高于传统谱分析方法的TWA阳性检测率,能够提高交替比率.TWA检测率明显高于physionet网站中2008年挑战（TWA检测和定量分析）得分第一的算法实验结果（66.67%）,说明该方法具有更强的TWA识别能力.

T-wave alternans detection based on enhanced spectral method and singular value decomposition

A method for the detection of T-wave alternans(TWA) was proposed through the fusion of enhanced spectral analysis and singular value decomposition in order to overcome the shortcoming of the traditional spectral analysis being sensitive to noise.Electrocardiosignal was effectively denoised by using singular value decomposition.The shortcomings of traditional spectral method including limitation to stationary signals and requirement of increased heart rates were overcome.The importance of alternans level was emphasized through increasing the amplitude of 0.5 cpb place on TWA spectrum in order to realize the effective analysis of TWA.Results showed that the accuracy of the algorithm was 93.33% for 30 synthetic electrocardiograms containing TWA in T-wave alternans database with increased alternans ratio and accuracy compared with traditional spectral analysis,which was obviously better than the result(66.67%) from the algorithm scored first in physionet 2008 challenge(TWA detection and quantitative analysis).The method is more effective for identifying TWA.