Self-Attention Generative Adversarial Network for Speech Enhancement
This work improves speech enhancement for applications like audio processing, but it is incremental as it builds on existing SEGAN with a novel attention mechanism.
The authors tackled the problem of speech enhancement by addressing the limitation of existing GANs that rely solely on convolution, which may obscure temporal dependencies, by proposing a self-attention layer integrated into a speech enhancement GAN (SEGAN) using raw signal input, resulting in consistent improvement across objective evaluation metrics.
Existing generative adversarial networks (GANs) for speech enhancement solely rely on the convolution operation, which may obscure temporal dependencies across the sequence input. To remedy this issue, we propose a self-attention layer adapted from non-local attention, coupled with the convolutional and deconvolutional layers of a speech enhancement GAN (SEGAN) using raw signal input. Further, we empirically study the effect of placing the self-attention layer at the (de)convolutional layers with varying layer indices as well as at all of them when memory allows. Our experiments show that introducing self-attention to SEGAN leads to consistent improvement across the objective evaluation metrics of enhancement performance. Furthermore, applying at different (de)convolutional layers does not significantly alter performance, suggesting that it can be conveniently applied at the highest-level (de)convolutional layer with the smallest memory overhead.