Xiao Yu

Guangwen Li, Xiao Yu

It is known belief propagation decoding variants of LDPC codes can be unrolled easily as neural networks after assigning differed weights to message passing edges flexibly. In this paper we focus on how to determine these weights, in the form of trainable paramters, within a framework of deep learning. Firstly, a new method is proposed to generate high-quality training data via exploiting an approximation to the targeted mixture density. Then the strong positive correlation between training loss and decoding metrics is fully exposed after tracing the training evolution curves. Lastly, for the purpose of facilitating training convergence and reducing decoding complexity, we highlight the necessity of slashing the number of trainable parameters while emphasizing the locations of these survived ones, which is justified in the extensive simulation.

Nameer Hirschkind, Xiao Yu, Mahesh Kumar Nandwana et al.

We introduce DiffuseST, a low-latency, direct speech-to-speech translation system capable of preserving the input speaker's voice zero-shot while translating from multiple source languages into English. We experiment with the synthesizer component of the architecture, comparing a Tacotron-based synthesizer to a novel diffusion-based synthesizer. We find the diffusion-based synthesizer to improve MOS and PESQ audio quality metrics by 23\% each and speaker similarity by 5\% while maintaining comparable BLEU scores. Despite having more than double the parameter count, the diffusion synthesizer has lower latency, allowing the entire model to run more than 5$\times$ faster than real-time.