Detection Of R-wave In Dynamic ECG Based On Quality Analysis

Nowadays, the electrocardiogram (ECG) was applied dynamic and in long-term monitoring in daily life, and the R-wave detection is the key in all kinds of ECG monitorings. Unfortunately, in the long term and dynamic recording, ECG signals may be involved in many kinds of noises. Electromyographic (EMG) interference is the most typical noise in dynamic ECG, and it has great influence on detection of R-wave.According to the experiments of dynamic ECG recording, features of dynamic ECG signals with heavy EMG artifact were analyzed. Different kinds of muscle movements produce different types of EMG, and there are generally two types of EMG in the dynamic ECG signals:one is electricity produced by muscular relaxation or contraction, and it has rapid amplitude changes and distributes mostly at low frequency; the other is produced by holding objects and keeping still, which has a huge distraction on R-wave and has more components at higher frequency than the first type.Therefore, based on the analysis of EMG interference, an R-wave detection method is proposed. We first demonstrate different changes of ECG morphology due to different types of EMGs. Then we classify EMG interference by their power ratio in frequency domain, zero-crossing rate and short time energy. Power ratio in frequency domain is used to classify EMG in different types with a long window size; the threshold-cross rate is used to classify the static ECG and dynamic ECG in a short window; the short time energy is used to classify the state of holding objects and swing arms. Combined with the results of frequency domain and time domain, the classification is reliable in short time.The choice of R-wave detection methods (wavelet transform and Empirical Mode Decomposition method) are related to the result of classification. Dmey wavelet transform is used for dynamic ECG of holding objects in multi-scale and multi-resolution. The tendency of ECG signals could be obtained by the EMD method which could decompose signals into a limited number of Intrinsic Mode Functions (IMF). The multi-scale analysis of IMFs is used to analyze the ECG of swing arms.The proposed adaptive R-wave detection method is validated using the data collected in our laboratory, and the experimental results demonstrate that the proposed method is significantly applicable in dynamic ECG.