Generative Speech Coding with Predictive Variance Regularization
This addresses speech compression for communication systems, offering a significant bit-rate reduction while maintaining quality, though it builds incrementally on existing generative modeling approaches.
The paper tackled the problem of generative speech codec performance deteriorating with real-world signal distortions, and introduced predictive-variance regularization to reduce sensitivity to outliers, resulting in state-of-the-art coding performance at 3 kb/s for real-world speech signals.
The recent emergence of machine-learning based generative models for speech suggests a significant reduction in bit rate for speech codecs is possible. However, the performance of generative models deteriorates significantly with the distortions present in real-world input signals. We argue that this deterioration is due to the sensitivity of the maximum likelihood criterion to outliers and the ineffectiveness of modeling a sum of independent signals with a single autoregressive model. We introduce predictive-variance regularization to reduce the sensitivity to outliers, resulting in a significant increase in performance. We show that noise reduction to remove unwanted signals can significantly increase performance. We provide extensive subjective performance evaluations that show that our system based on generative modeling provides state-of-the-art coding performance at 3 kb/s for real-world speech signals at reasonable computational complexity.