Real Time Speech Enhancement in the Waveform Domain

06/23/2020
by   Alexandre Défossez, et al.
0

We present a causal speech enhancement model working on the raw waveform that runs in real-time on a laptop CPU. The proposed model is based on an encoder-decoder architecture with skip-connections. It is optimized on both time and frequency domains, using multiple loss functions. Empirical evidence shows that it is capable of removing various kinds of background noise including stationary and non-stationary noises, as well as room reverb. Additionally, we suggest a set of data augmentation techniques applied directly on the raw waveform which further improve model performance and its generalization abilities. We perform evaluations on several standard benchmarks, both using objective metrics and human judgements. The proposed model matches state-of-the-art performance of both causal and non causal methods while working directly on the raw waveform.

READ FULL TEXT

page 1

page 2

page 3

page 4

research
02/15/2022

Speech Denoising in the Waveform Domain with Self-Attention

In this work, we present CleanUNet, a causal speech denoising model on t...
research
05/06/2021

DBNet: A Dual-branch Network Architecture Processing on Spectrum and Waveform for Single-channel Speech Enhancement

In real acoustic environment, speech enhancement is an arduous task to i...
research
06/08/2020

A non-causal FFTNet architecture for speech enhancement

In this paper, we suggest a new parallel, non-causal and shallow wavefor...
research
03/26/2023

Time-domain Speech Enhancement Assisted by Multi-resolution Frequency Encoder and Decoder

Time-domain speech enhancement (SE) has recently been intensively invest...
research
09/03/2020

Dense CNN with Self-Attention for Time-Domain Speech Enhancement

Speech enhancement in the time domain is becoming increasingly popular i...
research
07/03/2019

Convolutional Neural Network-based Speech Enhancement for Cochlear Implant Recipients

Attempts to develop speech enhancement algorithms with improved speech i...
research
07/11/2021

Neural Waveshaping Synthesis

We present the Neural Waveshaping Unit (NEWT): a novel, lightweight, ful...

Please sign up or login with your details

Forgot password? Click here to reset