Multi-Channel WSJ Audio

<h3>Introduction</h3> <p>Multi-Channel WSJ Audio (MCWSJ) was developed by the <a href="http://www.cstr.ed.ac.uk/" rel="nofollow">Centre for Speech Technology Research </a> at The University of Edinburgh and contains approximately 100 hours of recorded speech from 45 British English speakers. Participants read Wall Street Journal texts published in 1987-1989 in three recording scenarios: a single stationary speaker, two stationary overlapping speakers and one single moving speaker.</p> <p>This corpus was designed to address the challenges of speech recognition in meetings, which often occur in rooms with non-ideal acoustic conditions and significant background noise, and may contain large sections of overlapping speech. Using headset microphones represents one approach, but meeting participants may be reluctant to wear them. Microphone arrays are another option. MCWSJ supports research in large vocabulary tasks using microphone arrays. The news sentences read by speakers are taken from <a href="http://catalog.ldc.upenn.edu/LDC95S24" rel="nofollow">WSJCAM0 Cambridge Read News</a>, a corpus originally developed for large vocabulary continuous speech recognition experiments, which in turn was based on <a href="http://catalog.ldc.upenn.edu/LDC93S6A" rel="nofollow">CSR-1 (WSJ0) Complete</a>, made available by LDC to support large vocabulary continuous speech recognition initiatives. </p> <h3>Data</h3> <p> Speakers reading news text from prompts were recorded using a headset microphone, a lapel microphone and an eight-channel microphone array. In the single speaker scenario, participants read from six fixed positions. Fixed positions were assigned for the entire recording in the overlapping scenario. For the moving scenario, participants moved from one position to the next while reading. </p> <p>Fifteen speakers were recorded for the single scenario, nine pairs for the overlapping scenario and nine individuals for the moving scenario. Each read approximately 90 sentences. </p> <p>The audio data are presented as single channel 16kHz flac compressed wav files.</p> <h3>Samples</h3> <p>Please listen to the below samples.</p> <ul> <li><a href="./desc/addenda/LDC2014S03.olap.wav" rel="nofollow">Overlapping Sample</a></li> <li><a href="./desc/addenda/LDC2014S03.stat.wav" rel="nofollow">Stationary Sample</a></li> <li><a href="./desc/addenda/LDC2014S03.move.wav" rel="nofollow">Moving Sample</a></li> </ul> <h3>Updates</h3> <p>None at this time. </p> </br>
Portions © 1987-1989 Dow Jones & Company, Inc., © 2014 University Court of the University of Edinburgh, © 1992-1995, 2014 Trustees of the University of Pennsylvania

<h3>简介</h3> <p>多通道WSJ音频（Multi-Channel WSJ Audio, MCWSJ）由爱丁堡大学的<a href="http://www.cstr.ed.ac.uk/" rel="nofollow">语音技术研究中心（Centre for Speech Technology Research）</a>开发，包含来自45名英国英语使用者的约100小时录制语音。受试者朗读1987-1989年发布的《华尔街日报（Wall Street Journal）》文本，共三种录制场景：单固定站位讲话者、两名固定站位的重叠讲话者以及一名移动讲话者。</p> <p>该语料库旨在应对会议场景下的语音识别挑战——这类场景往往处于非理想声学环境，存在显著背景噪声，且可能包含大量重叠语音。使用头戴式麦克风是一种解决方案，但参会者通常不愿佩戴这类设备；麦克风阵列则是另一可选方案。MCWSJ可支持基于麦克风阵列的大词汇量语音识别研究。受试者朗读的新闻语句源自<a href="http://catalog.ldc.upenn.edu/LDC95S24" rel="nofollow">WSJCAM0剑桥朗读新闻语料库（WSJCAM0 Cambridge Read News）</a>，该语料库最初为大词汇量连续语音识别实验开发，其本身又基于<a href="http://catalog.ldc.upenn.edu/LDC93S6A" rel="nofollow">CSR-1（WSJ0）完整语料库（CSR-1 (WSJ0) Complete）</a>，由LDC发布以支持大词汇量连续语音识别相关研究。</p> <h3>数据</h3> <p>受试者根据提示朗读新闻文本，录制设备包括头戴式麦克风、领夹式麦克风以及八通道麦克风阵列。单讲话者场景下，受试者需在六个固定站位完成朗读；重叠场景下，受试者全程使用分配好的固定站位；移动场景下，受试者需边朗读边从一个站位移动至下一个站位。</p> <p>单场景录制了15名受试者，重叠场景录制了9组讲话者对，移动场景录制了9名受试者。每名受试者朗读约90个语句。</p> <p>音频数据以单通道16kHz FLAC压缩WAV文件格式提供。</p> <h3>示例</h3> <p>请收听以下示例音频。</p> <ul> <li><a href="./desc/addenda/LDC2014S03.olap.wav" rel="nofollow">重叠场景示例</a></li> <li><a href="./desc/addenda/LDC2014S03.stat.wav" rel="nofollow">固定站位场景示例</a></li> <li><a href="./desc/addenda/LDC2014S03.move.wav" rel="nofollow">移动场景示例</a></li> </ul> <h3>更新说明</h3> <p>暂无更新记录。</p> </br> <p>部分内容 © 1987-1989 道琼斯公司（Dow Jones & Company, Inc.），© 2014 爱丁堡大学校董会，© 1992-1995、2014 宾夕法尼亚大学理事会</p>