WHAM! (WSJ0 Hipster Ambient Mixtures)

“WSJ0 Hipster Ambient Mixtures (WHAM!) 数据集将 wsj0-2mix 数据集中的每个双说话者混合与独特的噪声背景场景配对。它有一个名为 WHAMR! 的扩展，它还为语音信号添加了人工混响到背景噪音。噪音音频于 2018 年底在旧金山湾区的各个城市地点收集。环境主要包括餐馆、咖啡馆、酒吧和公园。音频是使用三脚架上的 Apogee Sennheiser 双耳麦克风录制的离地 1.0 到 1.5 米。”

The WSJ0 Hipster Ambient Mixtures (WHAM!) dataset pairs each two-speaker mixture in the wsj0-2mix dataset with a unique noisy background scenario. It has an extension named WHAMR!, which adds artificial reverberation to the background noise for the speech signals. The noise audio was collected at various urban locations across the San Francisco Bay Area in late 2018. The environments mainly include restaurants, cafes, bars, and parks. The audio was recorded 1.0 to 1.5 meters above the ground using an Apogee Sennheiser binaural microphone mounted on a tripod.

搜集汇总

背景与挑战

背景概述

WHAM!数据集是一个用于音频处理的公开数据集，专注于语音分离和增强任务。它基于wsj0-2mix数据集，将双说话者混合语音与真实环境噪声背景结合，噪声采集自旧金山湾区的城市场所如咖啡馆和公园，使用专业双耳麦克风录制，以模拟嘈杂环境。该数据集适用于音源分离、语音分离和语音增强研究，发布于2019年，采用CC BY-NC 4.0许可证。