DiPCo

该数据集是第六届CHiME挑战赛的焦点，主要针对远距离多麦克风对话语音的自动标识和识别。该数据集包含了4位发言人的录音，使用了4个双耳麦克风和6个微软Kinect麦克风阵列设备进行录制。规模上，该数据集提供了16段对话，大约40小时的音频用于训练；同时，还提供了2段对话用于开发和评估，任务涵盖了说话人标识和语音识别。

This dataset is the focus of the 6th CHiME Challenge, targeting automatic speaker diarization and recognition of far-field multi-microphone conversational speech. The dataset contains recordings from 4 speakers, collected using 4 binaural microphones and 6 Microsoft Kinect microphone arrays. In terms of scale, the dataset provides 16 conversations (approximately 40 hours of audio) for training, and 2 additional conversations for development and evaluation. The covered tasks include speaker diarization and speech recognition.