An electrocardiogram (ECG) is an important signal in the detection of heart diseases. However, ordinary clinical ECG devices only record short-term signals, which are often insufficient for a complete diagnosis of heart diseases. To solve this problem, a long-term ECG monitoring system has been developed using modern technologies. This system allows remote monitoring of patients and the collection of ECG data, which is very important for early detection of heart diseases. In this study, a portable device for long-term monitoring of heart activity and a personal cloud platform for storing ECG data were created. The signals were filtered, and important features were extracted using the Wavelet Transform method to detect arrhythmias from the collected ECG data. The Symlet4 wavelet transform was used to filter and separate high-frequency, short signals (QRS complex and RR interval). The Coiflet2 wavelet transform was used to filter and characterize low-frequency, long-lasting signals (ST-segment, P-wave, T-wave, PR, and QT intervals). The wavelet threshold method was used for noise removal. Criteria such as signal-to-noise ratio (SNR), root mean square error (RMSE), and mean square error (MSE) were used to evaluate the quality of the cleaned signals. The results showed that The Symlet4 waveform performed better in detecting the QRS complex and RR interval—that is, the SNR was high, and the error was low. On the other hand, Coiflet2 performed better in detecting the ST-segment, T-wave, P-wave, PR-interval, and QT-interval—that is, both the SNR and RMSE were high for these parts. That is, each waveform performed better for certain parts of the cardiac signal: Symlet4 for fast and short signals, and Coiflet2 for slower and longer parts.