Table 3 - The classification accuracy of the proposed model on different arrhythmia Types Normal PAC PVC AF PAC and PVC PVC and AF Total Number of segments 42946 6414 8828 20764 3424 5512 87888 Accurately identified 39385 6072 8341 18965 3351 5013 81127 Accuracy (%) 91.71 94.67 94.48 91.34 97.87 90.95 92.31 quality started to decrease when the SNR decreased to 25 dB and became 0% when the SNR decreased to 10 dB. The black bars which represent the proportions of medium signal quality increased first and then decreased with the decrease of SNR. The red line represents the proportions of poor signal quality, and all the segments were classified as poor quality when the SNR decrease to -5 dB or lower. Muscle Artifact Stress Test: Fig. 4b shows the classification results from the muscle artifact stress test. The proportions of good signal quality started to decrease when the SNR decreased to 28 dB and became 0% when the SNR decreased to 8 dB. The red line represents the proportions of poor signal quality, and all the segments were classified as poor quality when the SNR decreased to -4 dB or lower. Electrode Motion Artifact Stress Test: Fig. 4c shows the classification results from the electrode motion artifact stress test. The proportions of good signal quality started to decrease when the SNR decreased to 28 dB and became 0% when the SNR decreased to -2 dB. The red line represents the proportions of poor signal quality, and all the segments were classified as poor quality when the SNR decreased to -7 dB or lower. Classification Results on ECG Segments with Arrhythmia The classification results of the model for ECG segments with PAC, PVC, AF, or mixed arrhythmia are listed in Table 3. The classification accuracy of the proposed model on different arrhythmia segments is higher than 90%, even if there were 2 Fig. 5. Example of a one-hour ECG recording of patient 02 in CPSC 2020 with marked signal quality. (a) Manually labeled; (b) Labeled by the proposed model; (c) and (d) ECG segments with good signal quality; (e) Medium signal quality; and (f) Poor signal quality. August 2022 IEEE Instrumentation & Measurement Magazine 47