The increasing demand for extensive and diverse electrocardiogram (ECG) datasets to support deep learning applications in clinical diagnostics is often hindered by the sensitive nature of patient data. Previous studies have demonstrated that augmenting datasets with robust synthetic data can improve the performance of deep learning models. This study introduces an Advanced-WaveGAN approach for generating synthetic 12-lead ECG signals to address the challenge of data scarcity. It utilizes a convolutional autoencoder to extract meaningful features from ECG signals, which are subsequently employed by the WaveGAN generator to produce high-quality synthetic ECG signals. A performance evaluation was conducted using various Advanced-WaveGAN configurations on the CODE-15% dataset. Results from the ablation study indicated that the optimized Advanced-WaveGAN architecture outperforms traditional WaveGAN models in terms of stability and loss metrics, providing an enhanced solution for generating realistic ECG data to improve the model performance of ECG applications.