Because the ECG contains much of information regarding various heart diseases, the interpretation of an ECG is essential for monitoring heart health. Sensors, data acquisition, and pre-processing, such as filtering and denoising, can be done on a single chip. In line with that, the interpretation technique is also being developed, related to the speed of calculations and accuracy. In general, there are four stages of processing ECG information, namely, pre-processing, QRS detection, feature extraction, and classification. This paper proposes another alternative for ECG beat classification, that eliminates the pre-processing stage and combines feature extraction and classification in a single calculation stage, namely ensemble MLP. This method is expected to reduce computational costs while maintaining accuracy of 97% or more and a large number of classes, with 10 or more.