Abstract Cardiovascular disease is one of the major diseases that threaten human health. Electrocardiogram (ECG) signal is an important indicator for the diagnosis of cardiovascular disease. Accurate analysis of ECG plays a key role in the diagnosis of cardiovascular disease. Underdeveloped areas have always been a high-risk area for cardiovascular disease and there are few doctors for diagnosing cardiovascular disease. One solution is using a telemedicine system for disease diagnosis. However, the quality of the ECG signal collected is not necessarily reliable and may impact diagnosis. In order to solve the problem, we have studied various methods for assessing the quality of ECG signals. In the paper, we analyzed the 12-lead ECG data provided by PhysioNet and selected two features of the time domain: the number of R peaks and the amplitude difference. These two features were extracted from the ECG data to form a matrix of 24 features. We trained the classification model with the feature matrix and achieved a classification accuracy of 95.80% on the test set. Experimental results demonstrated that the proposed Adaboost algorithm had advantages in accuracy compared with other algorithms for solving ECG quality assessment problems.