短时经验模态分解实时识别生猪心电QRS波群

心率是猪的重要生命体征,而在健康监测中由于猪的依从性较差而造成心电信号呈现非平稳特性,给实时心率计算带来困难。该文针对此问题结合经验模态分解方法(empirical mode decomposition,EMD)提出一种对心电信号具有实时处理能力的短时经验模态分解算法(short-time empirical mode decomposition,ST-EMD)。该算法通过对数据分段并根据信号特征决定分段起点及长度等参数,然后对每段数据进行EMD分解,再基于能量窗变换法从分解结果中提取QRS波的特征参数并识别R波。通过动物试验表明,ST-EMD算法能够对猪的心电信号实时处理和识别QRS波群,识别正确率为99.6%,且表现出一定的自适应性。说明本文提出的ST-EMD算法思路是正确的,适用于生猪的心电实时健康监护。

Short-time empirical mode decomposition method of real-time QRS complex identifying for electrocardiography of pig

Abstract: Pigs'' physiological parameters monitoring instantaneously in commercial piggeries is essential for early disease detection and welfare assessment. Heart rate is vital sign of pigs because of its correlation with disease and environment. Usually, the heart rate is calculated by the intervals of QRS complex identified from ECG (electrocardiogram) signal. Due to the low compliance of livestock during monitoring, the pig''s ECG signal presents non-stationary characteristics and varieties of noise, which makes it difficult to process instantaneously and correctly. To solve the problem, this study proposed a short-term empirical mode decomposition (ST-EMD) algorithm with the real-time and anti-noise property for ECG data processing based on empirical mode decomposition (EMD). The ST-EMD algorithm comprises 3 procedures including data segmenting, EMD processing and QRS complex feature extracting. In the data segmenting step, the algorithm determines the starting point and terminal point for the next data block according to the latest QRS complex and RRI (RR interval) in current data segment, and then captures the ECG data to meet the predetermined length in real time. After the signal collecting, the new fragment data are decomposed into a series of intrinsic mode functions (IMFs) and residuals by EMD algorithm. Through experiment analysis and comparison, the first IMF contains the most of QRS complex information. In the third procedure, the QRS complex features are extracted based on the first IMF employing energy window transformation and the correctness of the QRS complex identification is checked. After the ST-EMD algorithm was developed in MATLAB, animal experiments were carried out to verify the efficiency. Three piglets born in the same nest were recruited in the experiments. The age of the piglets was 50 d, and the average weight was 18.2 kg. Pig''s ECG signal was picked up by 2 electrodes attached on either side of chest and converted by BMD101 analog sensor with 512 Hz sampling frequency. Simultaneously, the software accessed the digital ECG data through Bluetooth communication and computed the heart rate in real time. The collection time of each pig was 120 s, and in total 184 320 data were collected. In the signal processing, the mean segmenting length of data block was 0.69 s, while the mean computing time was 0.03 s, which meant that the algorithm could output the heart rate data within 0.72 s after each ECG data block was acquired. Consequently, the ST-EMD algorithm can identify the QRS complex in real time from pig''s ECG signal. By comparing the actual QRS complexes, the mean accuracy of QRS complex identified by ST-EMD is 99.6%. The results of animal experiments illustrate that the proposed ST-EMD algorithm is correct and is suitable for the real-time health monitoring of the pig''s ECG. Moreover, to ensure the heart rate was correctly extracted by the algorithm, a seven-point moving median filtering was used to eliminate the erroneous values due to baseline drift and interference noise in ECG signal. As a result, the mean heart rates of 3 piglets were (137.47±6.47), (133.01±9.80) and (128±6.51) bmp respectively and all were within normal limits. In summary, the ST-EMD algorithm is efficacious and reliable for real-time processing of pig''s ECG signal integrated with moving median filtering.