由于P波一般为低频低幅波,容易受到基线漂移,肌电干扰等噪声影响,且不是每个心拍都包含P波,确定某一心拍有无P波也是一难题,针对小波-幅值-斜率的方法对多样形态P波适应的局限性,以及小波变换结合神经网络检测方法中选取伪P波样本的局限性,本文提出了基于小波-幅值阈值并以多特征作为神经网络的输入的P波检测方法,该方法首先利用小波变换对心电(ECG)信号进行去噪,然后利用小波变换求模极大值对的方法确定候选P波的位置,接下来利用幅值阈值初步判断有无P波,最后利用神经网络确定心拍有无P波。本文经由专家注释的QT心电数据库对该算法和传统的小波阈值法及基于小波-幅值-斜率的方法检测ECG信号P波的效果进行了对比,验证了本文提出的算法的可行性,对医院心电科记录的ECG信号进行了检测,其结果与医生的标注基本相同,并对QT数据库中的13份且每份15min的ECG信号进行了检测验证,P波正确检测率达到了99.911%。 Because P wave is generally a low-frequency low-amplitude wave, it is easily affected by noise such as baseline drift and electromyography interference, and not every heart beat includes P wave. It is also a difficult problem to determine whether a heart beat has P wave or not. Slope method to adapt to the various forms of P-wave and the limitations of the wavelet transform combined with neural network detection method to select the pseudo-P-wave samples, this paper presents a wavelet neural network based on the threshold value and multi-feature as the neural network input P wave detection method. Firstly, the ECG signal is denoised by using wavelet transform. Then, the position of the candidate P wave is determined by the method of wavelet transform modulus maxima. Then, No P wave, the last use of neural network to determine whether the heart beat P wave. In this paper, we compare the effectiveness of this algorithm with the traditional wavelet threshold method and P wave detection based on wavelet-amplitude-slope method by the expert annotation QT ECG database to verify the feasibility of the proposed algorithm, ECG signals recorded by the hospital’s cardiology department were tested and the results were basically the same as the doctors’ annotations. The 13 ECG signals of 15 minutes in the QT database were tested and verified. The correct detection rate of P wave reached 99.911% .