A DNN Based Post-Filter to Enhance the Quality of Coded Speech in MDCT Domain

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Gupta, Kishan, Korse, Srikanth, Edler, Bernd, Fuchs, Guillaume
Format: Preprint
Published: 2022
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916749971030016
author Gupta, Kishan
Korse, Srikanth
Edler, Bernd
Fuchs, Guillaume
author_facet Gupta, Kishan
Korse, Srikanth
Edler, Bernd
Fuchs, Guillaume
contents Frequency domain processing, and in particular the use of Modified Discrete Cosine Transform (MDCT), is the most widespread approach to audio coding. However, at low bitrates, audio quality, especially for speech, degrades drastically due to the lack of available bits to directly code the transform coefficients. Traditionally, post-filtering has been used to mitigate artefacts in the coded speech by exploiting a-priori information of the source and extra transmitted parameters. Recently, data-driven post-filters have shown better results, but at the cost of significant additional complexity and delay. In this work, we propose a mask-based post-filter operating directly in MDCT domain of the codec, inducing no extra delay. The real-valued mask is applied to the quantized MDCT coefficients and is estimated from a relatively lightweight convolutional encoder-decoder network. Our solution is tested on the recently standardized low-delay, low-complexity codec (LC3) at lowest possible bitrate of 16 kbps. Objective and subjective assessments clearly show the advantage of this approach over the conventional post-filter, with an average improvement of 10 MUSHRA points over the LC3 coded speech.
format Preprint
id arxiv_https___arxiv_org_abs_2201_12039
institution arXiv
publishDate 2022
record_format arxiv
spellingShingle A DNN Based Post-Filter to Enhance the Quality of Coded Speech in MDCT Domain
Gupta, Kishan
Korse, Srikanth
Edler, Bernd
Fuchs, Guillaume
Audio and Speech Processing
Machine Learning
Sound
Signal Processing
Frequency domain processing, and in particular the use of Modified Discrete Cosine Transform (MDCT), is the most widespread approach to audio coding. However, at low bitrates, audio quality, especially for speech, degrades drastically due to the lack of available bits to directly code the transform coefficients. Traditionally, post-filtering has been used to mitigate artefacts in the coded speech by exploiting a-priori information of the source and extra transmitted parameters. Recently, data-driven post-filters have shown better results, but at the cost of significant additional complexity and delay. In this work, we propose a mask-based post-filter operating directly in MDCT domain of the codec, inducing no extra delay. The real-valued mask is applied to the quantized MDCT coefficients and is estimated from a relatively lightweight convolutional encoder-decoder network. Our solution is tested on the recently standardized low-delay, low-complexity codec (LC3) at lowest possible bitrate of 16 kbps. Objective and subjective assessments clearly show the advantage of this approach over the conventional post-filter, with an average improvement of 10 MUSHRA points over the LC3 coded speech.
title A DNN Based Post-Filter to Enhance the Quality of Coded Speech in MDCT Domain
topic Audio and Speech Processing
Machine Learning
Sound
Signal Processing
url https://arxiv.org/abs/2201.12039