CHiME-5

CHiME-5是关于自动语音识别处理技术的数据集。该数据集来自第5个CHiME挑战，包括在真实家庭环境中进行远程多麦克风会话的任务。从晚餐场景中提取语音素材，数据集获得自然对话语音数据，并由6个Kinect麦克风阵列和4个双耳麦克风对记录。
数据集包含单阵列轨道与多阵列轨道语言建模，以及用于阵列同步，语音增强，常规和端到端ASR的数据收集过程，任务和基线系统。

CHiME-5 is a dataset for automatic speech recognition (ASR) processing technologies. Derived from the 5th CHiME Challenge, this dataset covers distant-talking multi-microphone session tasks conducted in real home environments. Speech materials are extracted from dinner scenarios, and natural conversational speech data is acquired and recorded via 6 Kinect microphone arrays and 4 binaural microphones.
The dataset includes language modeling for single-array tracks and multi-array tracks, as well as data collection procedures, tasks and baseline systems for array synchronization, speech enhancement, conventional and end-to-end ASR.