arXiv subjects

Zhaoxia Li

Publications and source records attributed to Zhaoxia Li.

3 recordsLinked to original sources

arXiv2023

Multi-Task Sub-Band Network For Deep Residual Echo Suppression

This paper introduces the SWANT team entry to the ICASSP 2023 AEC Challenge. We submit a system that cascades a linear filter with a neural post-filter. Particularly, we adopt sub-band processing to handle full-band signals and shape the network with multi-task learning, where dual signal voice activity detection (DSVAD) and echo estimation are adopted as auxiliary tasks. Moreover, we particularly improve the time frequency convolution module (TFCM) to increase the receptive field using small convolution kernels. Finally, our system has ranked 4th in ICASSP 2023 AEC Challenge Non-personalized track.

eess.AS↗

arXiv2022

The PCG-AIID System for L3DAS22 Challenge: MIMO and MISO convolutional recurrent Network for Multi Channel Speech Enhancement and Speech Recognition

This paper described the PCG-AIID system for L3DAS22 challenge in Task 1: 3D speech enhancement in office reverberant environment. We proposed a two-stage framework to address multi-channel speech denoising and dereverberation. In the first stage, a multiple input and multiple output (MIMO) network is applied to remove background noise while maintaining the spatial characteristics of multi-channel signals. In the second stage, a multiple input and single output (MISO) network is applied to enhance the speech from desired direction and post-filtering. As a result, our system ranked 3rd place in ICASSP2022 L3DAS22 challenge and significantly outperforms the baseline system, while achieving 3.2% WER and 0.972 STOI on the blind test-set.

eess.AS↗

arXiv2007

Gas Bubbles and Gas Pancakes at Liquid/Solid Interface: A Continuum Theory Incorporated with Molecular Interactions

The states of gas accumulated at the liquid-solid interface are analyzed based on the continuum theory where the Hamaker constant is used to describe the long-range interaction at the microscopic scale. The Hamaker constant is always negative, whereas the gas spreading coefficient can be either sign. Despite the complexity of gas, including that the density profile may not be uniform due to absorption on both solid and liquid surfaces, we predict three possible gas states at the liquid-solid interface, i.e. complete wetting, partial wetting and pseudopartial wetting. These possible gas states correspond respectively to a gas pancake (or film) surrounded by a wet solid, a gas bubble with a finite contact angle, and a gas bubble(s) coexisting with a gas pancake. Typical thickness of the gas pancakes is at the nanoscale within the force range of the long-range interaction, whereas the radius of the gas bubbles can be large. The state of gas bubble(s) coexisting with a gas film is predicted theoretically for the first time. Our theoretical results can contribute to the development of a unified picture of gas nucleation at the liquid-solid interface.

cond-mat.soft↗