The major function of heart is to pump blood to tissues and organs necessary for the body metabolism. It is therefore one of the organs that affects human life. However, adverse situations, such as paralysis and death are the major problems that can lead to a heart failure. Healthy heart is very important to live comfortably. To prevent adverse events, it is important to monitor and detect heart diseases early. The aim of proposed method is to determine and classify nine types of ECG arrhythmias, including normal beats. A large feature set was obtained from the MIT-BIH Arrhythmia database. Zhao Atlas-Mark time-frequency distribution was used to extract the feature set. Five classification algorithms have been tried. The Cubic Support Vector Machine algorithm yielded best performance results. The proposed method achieved accuracy, sensitivity, specificity, F-score, positive predictive, and negative predictive values of 96.39%, 94.22%, 92.02%, 93.91%, 93.90% and 96.72%, respectively. Considering the data size, performance values, and number of arrythmias, the proposed method provided superiority to other studies. Furthermore, running time is suitable for telemedicine systems.