VOICES (Voices Obscured In Complex Environmental Settings)

VOICES 语料库是一个数据集，用于促进对远场麦克风在嘈杂的房间条件下录制的语音进行语音和信号处理研究。对于这个语料库，音频是在带家具的房间中录制的，背景噪音与从 LibriSpeech 语料库中选择的前景语音一起播放。每个房间都记录了多个会话，以适应所有前景语音-背景噪声组合。使用放置在整个房间的 12 个麦克风录制音频，每个麦克风可录制 120 小时的音频。

The VOICES Corpus is a dataset designed to advance research on speech and signal processing for speech recorded by far-field microphones under noisy room conditions. For this corpus, audio was captured in furnished rooms, with background noise paired with foreground speech selected from the LibriSpeech corpus. Multiple recording sessions were conducted for each room to cover all combinations of foreground speech and background noise. Audio was recorded using 12 microphones placed throughout the room, with each microphone capable of recording 120 hours of audio.

搜集汇总

背景与挑战

背景概述

VOICES语料库是一个用于语音和信号处理研究的数据集，专注于在嘈杂房间条件下通过远场麦克风录制的语音。它包含在带家具环境中使用12个麦克风录制的120小时音频，结合了背景噪音和LibriSpeech语料库的前景语音，于2018年发布。