Cardiovascular diseases, especially atrial fibrillation (AFib), are the leading causes of morbidity and mortality worldwide. Timely and accurate detection of AFib from electrocardiogram (ECG) signals remains a challenge, mainly due to the episodic nature and often asymptomatic presentation of AFib. This paper introduces a novel hybrid deep learning model combining Convolutional Neural Networks (CNN) and Bidirectional Long Short-Term Memory (Bi-LSTM) networks to enhance ECG signal classification. Unlike existing models, our approach integrates advanced signal processing techniques such as spectral entropy and instantaneous frequency estimation to enhance feature extraction further. Experimental results demonstrate that our model achieves a classification accuracy of 96%, significantly outperforming baseline models, including CNN-LSTM and traditional LSTM, with improvements in generalization and robustness. This architecture can be applied in real-time diagnostic tools for early AFib detection, improving cardiovascular disease management.