The electrocardiogram (ECG) remains crucial in diagnosing cardiac arrhythmias due to its widespread use and cost-effectiveness. Analyzing the morphological P-wave features offers vital insights into abnormalities in interatrial and atrioventricular (AV) conduction, essential for identifying atrial arrhythmias. However, discriminating different types of atrial arrhythmias can be challenging and time-consuming. Standard 12-lead ECG systems often struggle to detect subtle P-wave abnormalities, leading to increased misdiagnosis. Atrial arrhythmias, including Atrial Fibrillation (AF), Atrial Flutter (AFL), and Atrial Tachycardia (AT), are prevalent among hospitalized individuals, significantly impacting morbidity and mortality rates. Hence, enhancing ECG signal quality is essential for minimizing false positives and improving diagnostic accuracy in identifying atrial arrhythmias. This thesis aimed to enhance atrial arrhythmia detection by strengthening atrial activity signal (P/f/F-waves) through optimal lead modification and automated algorithm implementation. Machine Learning (ML) and Deep Learning (DL) algorithms, powered by Artificial Intelligence (AI), optimized lead selection and accurately classified arrhythmias, facilitating improved detection and diagnosis. A novel Atrial Lead System (ALS) enhanced P-wave signal strength, while advanced Gradient Boosting (GB) and DL algorithms ranked optimal bipolar leads. CatBoost ML model automated lead selection improved P-wave change detection, outperforming other ML models. ALS improved signal strength, noise resistance, and reliability, especially for mobile health (mHealth) applications. An automatic classification technique employing a 1D-CNN-BiLSTM model for effectively distinguishing atrial arrhythmias from normal Sinus Rhythm (SR) achieves 94% accuracy in both the cross-validation and testing phases. These findings emphasize ALS&rsquos efficacy in enhancing P-wave signal strength and AV ratios, advocating its integration into clinical devices for accurate atrial arrhythmia diagnosis in real-time settings.