RWCP-SSD-Onomatopoeia

RWCP-SSD-Onomatopoeia是一个由立命馆大学等机构创建的数据集，包含155,568个拟声词及其对应的音频样本，用于环境声音合成。数据集涵盖了105种声音事件，每个事件包含约100个音频样本，总样本数达9,722个。创建过程中，通过众包方式收集了拟声词，并记录了众包工人的自我报告信心分数和他人报告的接受分数。该数据集旨在解决环境声音合成中对声音精细时间-频率结构控制的需求，适用于电影、游戏制作、虚拟现实内容生成及声音事件检测和场景分类的数据增强。RWCP-SSD-Onomatopoeia is a dataset developed by institutions including Ritsumeikan University, which contains 155,568 onomatopoeic annotations and their corresponding audio samples for environmental sound synthesis. The dataset covers 105 sound event categories, with approximately 100 audio samples per category, amounting to a total of 9,722 audio samples. During the dataset creation process, onomatopoeic annotations were collected via crowdsourcing, and both self-reported confidence scores from crowdworkers and peer-reported acceptance scores were recorded. This dataset aims to address the demand for fine-grained time-frequency structure control in environmental sound synthesis, and is suitable for data augmentation in film production, game development, virtual reality content generation, as well as sound event detection and scene classification tasks.

搜集汇总

背景与挑战

背景概述

RWCP-SSD-Onomatopoeia是一个专注于环境声音合成的数据集，包含155,568个拟声词和9,722个音频样本，覆盖105种声音事件，通过众包方式收集并附带信心和接受分数。该数据集旨在提供对声音精细时间-频率结构的控制，适用于影视游戏制作、虚拟现实内容生成以及声音事件检测的数据增强。