Massive Distributed Microphone Array Dataset

Large, distributed microphone arrays could offer dramatic advantages for audio source separation, spatial audio capture, and human and machine listening applications. This dataset contains acoustic measurements and speech recordings from 10 loudspeakers and 160 microphones spread throughout a large, reverberant conference room. The distributed microphone system contains two types of array: four wearable microphone arrays of 16 sensors each placed near the ears and across the upper body, and twelve tabletop arrays of 8 microphones each in enclosures designed to resemble voice-assistant speakers. The dataset includes recordings of chirps that can be used to measure impulse responses and of speech clips derived from the CSTR VCTK corpus. The speech clips are recorded both individually and as a mixture to support source separation experiments. The uncompressed files are about 13.4 GB.大型分布式麦克风阵列在音频源分离、空间音频采集以及人机听觉应用中具备显著优势。本数据集包含来自10个扬声器与160个麦克风的声学测量数据与语音录音，采集设备布设于一间大型混响会议室的各处。该分布式麦克风系统包含两类阵列：一是4个可穿戴麦克风阵列，每个阵列集成16个传感器，佩戴于耳部附近及上身区域；二是12个桌面式麦克风阵列，每个阵列配备8个麦克风，其外壳设计仿照智能语音助手音箱。本数据集包含可用于测量脉冲响应的啁啾信号录音，以及源自CSTR VCTK语料库（CSTR VCTK corpus）的语音片段。这些语音片段既包含单一声源独立录制版本，也包含多声源混合录制版本，可用于支撑音频源分离相关实验。未压缩文件总容量约为13.4 GB。

搜集汇总

背景与挑战

背景概述

该数据集是一个大规模分布式麦克风阵列数据集，包含10个扬声器和160个麦克风，分布在大型混响会议室中，涵盖可穿戴和桌面两种阵列类型，用于音频源分离、空间音频捕获等研究。数据包括啁啾声和语音片段，支持源分离实验，未压缩文件约13.4 GB，发布于2019年，适用于技术与工程领域。