This thesis presents a novel method on a Smart-phone for ECG tele-monitoring signal analysis. The proposed system focuses on QRS complex detection, beat classification and arrhythmias classification. In the regular process, the QRS complex is detected by the Pan-Tompkins algorithm and classified as normal sinus rhythms (SRs) or pre- mature ventricular contractions (PVCs) by existing classification methods. Subse- quently, the Lempel-Ziv (LZ) complexity measure, including the K-Means clustering algorithm and the LZ complexity analysis, is utilized to further separate the high risk arrhythmias, ventricular tachycardia (VT) or ventricular fibrillation (VF). In this procedure of the high risk arrhythmias, three consecutive PVC beats in a row are considered to be an indication of the beginning of VT rhythms, at which point the following data points will be saved until up to a certain window length long are reached. The window length long ECG signal will be further classified as VT or VF by several new decision rules with heart rate detection. Furthermore, the proposed system successfully implemented on a Smart-phone adopts the time frames to indicate the analysis report for improving the reliability and error detection of arrhythmias. The new analysis method presents fairly good performance results when applied to testing records chosen from the MIT-BIH database.