EasyCom
收藏资源简介:
EasyCom数据集是由Facebook Reality Labs Research创建的增强现实数据集,专注于在嘈杂环境中改善通信。该数据集包含超过5小时的多模态数据,适用于训练和测试信号处理及机器学习算法,如波束形成和语音增强。数据集内容包括AR眼镜的多通道麦克风阵列音频、广角RGB视频、语音源姿态、头戴式麦克风音频、标注的语音活动、语音转录、头部边界框、语音目标和源识别标签。创建此数据集旨在促进多模态AR解决方案在解决鸡尾酒会问题方面的研究。
The EasyCom dataset is an augmented reality (AR) dataset created by Facebook Reality Labs Research, focusing on improving communication in noisy environments. It contains over 5 hours of multimodal data, suitable for training and testing signal processing and machine learning algorithms such as beamforming and speech enhancement. The dataset includes multi-channel microphone array audio from AR glasses, wide-angle RGB video, speech source poses, headset microphone audio, annotated speech activity, speech transcripts, head bounding boxes, and speech target and source identification labels. This dataset was developed to promote research on multimodal AR solutions for solving the cocktail party problem.




