WaveNetEQ — Packet Loss Concealment with WaveRNN
By Florian Stimberg, Alex Narest, A. Bazzica, Lennart Kolmodin, Pablo González, Olga Sharonova, H. Lundin, Thomas C. Walters. 2020
17 citations recorded by the source corpus.
We present WaveNetEQ, a novel packet loss concealment method based on a WaveRNN architecture. The model is conditioned on a log-mel spectrogram of the past signal to extract slow moving features, like voice characteristics and prosody and achieves significantly better quality than pattern based methods for medium and long term packet loss. Through aggressive sparsification the model is efficient enough to run on a phone.
Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Speech & Audio, Signal Processing
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Marco Tagliasacchi
- Neil Zeghidour
- Alejandro Luebs
- Jan Skoglund
- Ahmed Omran
- Yossi Adi
- Zalán Borsos
- Hubert Siuzdak
In Computer Vision and Pattern Recognition
In Speech & Audio
- Marco Tagliasacchi
- Neil Zeghidour
- Jan Skoglund
- Ahmed Omran
- Alejandro Luebs
- Zalán Borsos
- Hubert Siuzdak
- Yossi Adi
In Signal Processing
- Marco Tagliasacchi
- Neil Zeghidour
- Jan Skoglund
- Alejandro Luebs
- Ahmed Omran
- Yossi Adi
- Zalán Borsos
- Hubert Siuzdak
Citations in the corpus
- SoundStream: An End-to-End Neural Audio Codec (2021, background): “This can be accomplished either via audio superresolution, i.e., extending the frequency bandwidth [1], via audio denoising, i.e., removing lossy coding artifacts [2], or via packet loss concealment [3].”