In this paper we propose a lightweight model for frequency bandwidth extension of speech signals, increasing the sampling frequency from 8kHz to 16kHz while restoring the high frequency content to a level almost indistinguishable from the 16kHz ground truth. The model architecture is based on SEANet (Sound EnhAncement Network), a wave-to-wave fully convolutional model, which uses a combination of feature losses and adversarial losses to reconstruct an enhanced version of the input speech. In addition, we propose a variant of SEANet that can be deployed on-device in streaming mode, achieving an architectural latency of 16ms. When profiled on a single core of a mobile CPU, processing one 16ms frame takes only 1.5ms. The low latency makes it viable for bi-directional voice communication systems.
@article{arxiv.2010.10677,
title = {Real-time Speech Frequency Bandwidth Extension},
author = {Yunpeng Li and Marco Tagliasacchi and Oleg Rybakov and Victor Ungureanu and Dominik Roblek},
journal= {arXiv preprint arXiv:2010.10677},
year = {2021}
}