本文介绍ConferencingSpeech2021挑战，聚焦远场复杂环境下的多通道语音增强，涉及单麦克风阵列、多分布麦克风阵列等任务。参与者将使用真实场景录音，通过主观评价和ACR/MOS评估系统进行竞赛。奖品丰厚，优秀者有机会获得实习机会。挑战包括两个任务：单麦克风阵列处理和多麦克风分布式处理。
Introduction
Welcome to Far-field Multi-channel Speech Enhancement Challenge for Video Conferencing (ConferencingSpeech 2021). With the advances in video conferencing technology, we are able to seamlessly connect with people of our choice anytime anywhere in the world. Video conferencing helps break barriers of distance among people. However, during video conference, the speech quality will be significantly affected by background noise, reverberation, number of recording microphones, the layout of microphone array, the acoustic and circuit design of microphone arrays, and so on. Effective speech enhancement plays an important role in the video conferencing system. Although the performance of speech enhancement has been improved dramatically in the past several decades, there are still a set of open research problems that should be further addressed in the far-field and complex meeting room environments, which includes but not limited to:
● Multi-channel speech enhancement with single microphone array
● Multi-channel speech enhancement with noisy and reverberant environments
● Multi-channel speech enhancement with multiple distributed microphone arrays
● Multi-channel speech enhancement with low-latency and zero look-ahead (casual system)
ConferencingSpeech 2021 challenge is proposed to stimulate research in the areas mentioned above. Targeting the real video conferencing room application, the challenge database is recorded from real speakers. Number of speakers and distances between speakers and microphone arrays vary according to the sizes of meeting rooms. Multiple microphone arrays with different geometric topologies are allocated in each recording environment. The challenge will provide the list of training set, development set, scripts for simulation, and baseline systems for participants to develop their systems. The final ranking of the challenge will be decided by the subjective evaluation. The subjective evaluation will be performed using Absolute Category Ratings (ACR) to estimate a Mean Opinion Score (MOS) through Tencent Online Media Subjective Evaluation platform.
The challenge has two tasks:
Task 1
Multi-channel speech enhancement with single microphone array
Task 2
Multi-channel speech enhancement with multiple distributed microphone arrays
Each registered team could participate in any one or both tasks. Participating in both tasks are encouraged. Top two winning teams from each task will be awarded the prizes provided by Tencent Ethereal Audio Lab. The First Prize is 1500 USD. The Second Prize is 800 USD. The internship opportunities at Tencent Ethereal Audio Lab will be provided to the students with excellent performance.More details about the data and ConferencingSpeech 2021 challenge can be found here.
Timeline
| 22 Jan 2021 | Release of evaluation plan, the list of training data, noise dataset, development test set, and scripts for simulation |
| 05 Feb 2021 | Release of baseline system |
| 07 Mar 2021 | Deadline of challenge registration |
| 08 Mar 2021 | Release of evaluation test set |
| 16 Mar 2021 | Deadline of submitting the results for subjective evaluation on the evaluation test set |
| 25 Mar 2021 | Notification of the results of participants |
| 26 Mar 2021 | Interspeech paper submission deadline |
| 02 Apr 2021 | Interspeech paper update deadline |
| 02 Jun 2021 | Paper acceptance/rejection notification |
| 05 Jun 2021 | Notification of the winners |
| 31 Aug 2021 | Opening of Interspeech 2021 |
Download
Participants will receive the download link/password when their registrations are confirmed.
Organizers
Contact Us
If you have questions about this challenge, please email us at ConferencingSpeech@tencent.com. If you haven’t received the reply in three days, please send email to ConferencingSpeech@gmail.com.
ConferencingSpeech 2021 Website
扫一扫
1080
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
