0%

Dual-Path Magnitude-Phase Learning with Bidirectional Cross-Attention for Bone-Conducted Speech Restoration

Posted on 2026-04-27 Edited on 2026-04-28 In Speech Enhancement

Author: Dianzhe Ding
Email: dingdianzhe@mail.ioa.ac.cn

Introduction

We present some speech samples in the website to show the proposed dual-path bone-conducted speech restoration network. Objective and subjective evaluations on A4BS dataset show that our system substantially outperforms existing bone-conducted speech enhancement systems.

Samples

We encourage our readers to listen to the following audio samples in order to experience the audio quality of our enhanced speech. Due to the large number of audio files, the loading time might be prolonged. Kindly wait for a moment or consider closing other web pages and refreshing the website to improve the loading speed.

Sample1	AC	BC	EBEN	DPT-EGNet
	WaveNet	Two Stage Approach	Tramba	TFiLM
	Tunet	MP-SENet	U-former	Proposed Network

Sample2	AC	BC	EBEN	DPT-EGNet
	WaveNet	Two Stage Approach	Tramba	TFiLM
	Tunet	MP-SENet	U-former	Proposed Network

Sample3	AC	BC	EBEN	DPT-EGNet
	WaveNet	Two Stage Approach	Tramba	TFiLM
	Tunet	MP-SENet	U-former	Proposed Network

Sample4	AC	BC	EBEN	DPT-EGNet
	WaveNet	Two Stage Approach	Tramba	TFiLM
	Tunet	MP-SENet	U-former	Proposed Network

Sample5	AC	BC	EBEN	DPT-EGNet
	WaveNet	Two Stage Approach	Tramba	TFiLM
	Tunet	MP-SENet	U-former	Proposed Network

Sample6	AC	BC	EBEN	DPT-EGNet
	WaveNet	Two Stage Approach	Tramba	TFiLM
	Tunet	MP-SENet	U-former	Proposed Network

Sample7	AC	BC	EBEN	DPT-EGNet
	WaveNet	Two Stage Approach	Tramba	TFiLM
	Tunet	MP-SENet	U-former	Proposed Network

Sample8	AC	BC	EBEN	DPT-EGNet
	WaveNet	Two Stage Approach	Tramba	TFiLM
	Tunet	MP-SENet	U-former	Proposed Network

Sample9	AC	BC	EBEN	DPT-EGNet
	WaveNet	Two Stage Approach	Tramba	TFiLM
	Tunet	MP-SENet	U-former	Proposed Network

Sample10	AC	BC	EBEN	DPT-EGNet
	WaveNet	Two Stage Approach	Tramba	TFiLM
	Tunet	MP-SENet	U-former	Proposed Network

Sample11	AC	BC	EBEN	DPT-EGNet
	WaveNet	Two Stage Approach	Tramba	TFiLM
	Tunet	MP-SENet	U-former	Proposed Network

Sample12	AC	BC	EBEN	DPT-EGNet
	WaveNet	Two Stage Approach	Tramba	TFiLM
	Tunet	MP-SENet	U-former	Proposed Network

Sample13	AC	BC	EBEN	DPT-EGNet
	WaveNet	Two Stage Approach	Tramba	TFiLM
	Tunet	MP-SENet	U-former	Proposed Network

Sample14	AC	BC	EBEN	DPT-EGNet
	WaveNet	Two Stage Approach	Tramba	TFiLM
	Tunet	MP-SENet	U-former	Proposed Network

Sample15	AC	BC	EBEN	DPT-EGNet
	WaveNet	Two Stage Approach	Tramba	TFiLM
	Tunet	MP-SENet	U-former	Proposed Network

Sample16	AC	BC	EBEN	DPT-EGNet
	WaveNet	Two Stage Approach	Tramba	TFiLM
	Tunet	MP-SENet	U-former	Proposed Network

Sample17	AC	BC	EBEN	DPT-EGNet
	WaveNet	Two Stage Approach	Tramba	TFiLM
	Tunet	MP-SENet	U-former	Proposed Network

Sample18	AC	BC	EBEN	DPT-EGNet
	WaveNet	Two Stage Approach	Tramba	TFiLM
	Tunet	MP-SENet	U-former	Proposed Network

1. Introduction
2. Samples

IOA-Audio

Audio & Speech Research Group is a research group under the Key Laboratory of Noise and Vibration Research, Institute of Acoustics, Chinese Academy of Sciences.
The Principal Investigator of this research group is Prof. Feiran Yang (https://people.ucas.ac.cn/~feiran). Our research interests include Blind Source Separation, Array Signal Processing, Voice Anti-Spoofing, Acoustic Echo Cancellation, Speech Enhancement. Contact us: feiran@mail.ioa.ac.cn

GitHub E-Mail