Source-Aware Neural Audio Coding

Paper

Haici Yang, Kai Zhen, Seungkwon Beack, and Minje Kim, “Source-Aware Neural Speech Coding for Noisy Speech Compression,” in Proceedings of the IEEE International Conference on Acoustics, Speech, and Signal Processing (ICASSP), Toronto, Canada, June 6-12, 2021. [pdf, presentation video]

Sound Examples

9.14 kbps

Jungle noise

Input mixture

Baseline (mixture)

SANAC (mixture)

Baseline (speech)

SANAC (speech)

Casino noise

Input mixture

Baseline (mixture)

SANAC (mixture)

Baseline (speech)

SANAC (speech)

Motorcycle

Input mixture

Baseline (mixture)

SANAC (mixture)

Baseline (speech)

SANAC (speech)

18.29 kbps

Jungle noise

Input mixture

Baseline (mixture)

SANAC (mixture)

Baseline (speech)

SANAC (speech)

Casino noise

Input mixture

Baseline (mixture)

SANAC (mixture)

Baseline (speech)

SANAC (speech)

Motorcycle

Input mixture

Baseline (mixture)

SANAC (mixture)

Baseline (speech)

SANAC (speech)

27.43 kbps

Jungle noise

Input mixture

Baseline (mixture)

SANAC (mixture)

Baseline (speech)

SANAC (speech)

Casino noise

Input mixture

Baseline (mixture)

SANAC (mixture)

Baseline (speech)

SANAC (speech)

Motorcycle

Input mixture

Baseline (mixture)

SANAC (mixture)

Baseline (speech)

SANAC (speech)

Source code

https://github.com/haiciyang/SANAC