ICASSP 2021 Acoustic Echo Cancellation Challenge Datasets

ICASSP 2021 Acoustic Echo Cancellation Challenge数据集由微软公司创建，旨在推动声学回声消除（AEC）领域的研究。该数据集包含超过2,500个真实音频设备和人类发言者的录音，涵盖单讲和双讲场景，以及一个合成数据集。数据集通过大规模众包方式收集，包含多种环境和设备的真实录音，以及模拟房间脉冲响应和背景噪声的合成数据。该数据集适用于深度学习和实际通信产品中的应用，旨在解决音频通信和会议系统中的回声问题，提高语音质量。

The ICASSP 2021 Acoustic Echo Cancellation (AEC) Challenge Dataset was developed by Microsoft Corporation to advance research in the field of acoustic echo cancellation. Comprising over 2,500 recordings of real audio devices and human speakers, the dataset covers both single-talk and double-talk scenarios, as well as a standalone synthetic dataset. Collected via large-scale crowdsourcing, the dataset includes real recordings from diverse environments and devices, alongside synthetic data that simulates room impulse responses and background noise. This dataset is applicable to deep learning and real-world communication product applications, aiming to resolve echo issues in audio communication and conferencing systems and improve speech quality.