CBLMA: A Hybrid Model for Robust Signal Classification with Convolutional and Bidirectional LSTM Layers

The accurate categorisation of modulation schemes under different signal-to-noise ratios (SNRs) has become a significant and challenging issue in the context of rapidly evolving and increasingly complex contemporary communication systems. This paper addresses this significant challenge by proposing a novel and sophisticated hybrid model, namely, Convolutional Bidirectional Long Short-Term Memory (CBLMA) with Multiple Heads of Attention. The methodology adopted in this research project involves the seamless and harmonious integration of convolutional layers for the extraction of detailed spatial features, bi-directional LSTM layers dedicated to the capture of complex temporal dependencies, and a multi-head attention mechanism that plays a key role in the dynamic and accurate adjustment of focus. The experimental results clearly illustrate the model's exceptional performance in classifying simple modulation schemes. Furthermore, it demonstrates considerable resilience and adaptability in addressing intricate modulation scenarios and high-noise conditions. It is notable that the CBLMA model offers a highly practical and efficient solution to the signal classification task while simultaneously facilitating the exploration of a multitude of promising avenues for a diverse array of applications within the expansive domain of communication systems.