DIRHA English WSJ Audio

<h3>Introduction</h3><br>
<p>DIRHA English WSJ Audio was developed as part of the <a href="https://dirha.fbk.eu/">Distant-Speech Interaction for Robust Home Applications (DIRHA) Project</a> which addressed natural spontaneous speech interaction with distant microphones in a domestic environment. It is comprised of approximately 85 hours of real and simulated read speech by six native American English speakers. The target utterances were taken from CSR-I (WSJ0) Complete (<a href="../../../LDC93S6A/">LDC93S6A</a>), specifically, the 5,000 word subset of read speech from Wall Street Journal news text.</p><br>
<p>This release contains signals of different characteristics in terms of noise and reverberation making it suitable for various multi-microphone signal processing and distant speech recognition tasks. The corpus can be coupled with related Kaldi baselines and tools that are available <a href="https://github.com/SHINE-FBK/DIRHA_English_wsj">here</a>.</p><br>
<h3>Data</h3><br>
<p>Speech was collected in a real apartment setting with typical domestic background noise and inter/intra-room reverberation effects. A total of 32 microphones were placed in the living-room (26 microphones) and in the kitchen (6 microphones). The original recordings were made at a sampling frequency of 48 kHz. However, for the sake of compactness, the released signals in this publication are in wav format with 16 kHz sampling frequency and 16 bit resolution.</p><br>
<p>Annotations for each acoustic sequence are included in xml format, such as microphone positions, speaker id, speaker gender and speaker position. Additional metadata about the speakers and images of the apartment setting are also provided. Consult the documentation accompanying this release for more information about the collection.</p><br>
<h3>Samples</h3><br>
<p>Please view this <a href="desc/addenda/LDC2018S01.wav">audio sample</a>&nbsp;and <a href="desc/addenda/LDC2018S01.xml">annotation sample</a>.</p><br>
<h3>Updates</h3><br>
<p>None at this time.</p></br>
Portions © 1987-1989 Dow Jones & Company, Inc., © 2018 Fondazione Bruno Kessler, © 1996, 2018 Trustees of the University of Pennsylvania

<h3>引言</h3><br><p>DIRHA英语WSJ语音数据集是面向稳健家庭应用的远场语音交互（Distant-Speech Interaction for Robust Home Applications, DIRHA）项目的研究成果，该项目旨在研究家庭环境下基于远场麦克风的自然自发语音交互。本数据集包含由6名以美式英语为母语的说话人录制的真实与模拟朗读语音，总时长约85小时。目标语句取自CSR-I（WSJ0）完整语料库（LDC93S6A），具体为《华尔街日报》新闻文本朗读语音中的5000词子集。</p><br><p>本次发布的数据集包含不同噪声与混响特性的语音信号，适用于多麦克风信号处理及远场语音识别等多种任务。该语料库可搭配相关Kaldi基线模型与工具使用，相关资源可参见<a href="https://github.com/SHINE-FBK/DIRHA_English_wsj">此处</a>。</p><br><h3>数据</h3><br><p>语音数据采集于真实公寓场景，包含典型的家庭背景噪声以及房间内/房间间混响效应。共在客厅（26个麦克风）与厨房（6个麦克风）部署了32个麦克风。原始录制采样率为48 kHz，但为压缩体积，本次发布的语音信号采用16 kHz采样率、16比特量化精度的WAV格式。</p><br><p>每条语音序列的标注均以XML格式提供，内容包含麦克风位置、说话人ID、说话人性别及说话人位置。此外还附带了说话人相关元数据与公寓场景的图像资料。如需了解数据采集的更多细节，请参阅本次发布包附带的说明文档。</p><br><h3>样本</h3><br><p>请查看<a href="desc/addenda/LDC2018S01.wav">语音样本</a>与<a href="desc/addenda/LDC2018S01.xml">标注样本</a>。</p><br><h3>更新记录</h3><br><p>暂无更新记录。</p></br>部分内容 © 1987-1989 Dow Jones & Company, Inc., © 2018 Fondazione Bruno Kessler, © 1996, 2018 Trustees of the University of Pennsylvania

搜集汇总

背景与挑战

背景概述

DIRHA English WSJ Audio是一个专注于家庭环境远场语音交互的英语语音数据集，包含约85小时由六位美国英语母语者录制的真实和模拟朗读语音，语料源自华尔街日报文本。数据在真实公寓中采集，使用32个麦克风，涵盖噪声和混响效应，提供XML格式的详细注释（如麦克风位置和说话者信息），适用于多麦克风信号处理和远场语音识别研究。