[DCASE2022 Task 3] Synthetic SELD mixtures for baseline training

DESCRIPTION: This audio dataset serves serves as supplementary material for the DCASE2022 Challenge Task 3: Sound Event Localization and Detection Evaluated in Real Spatial Sound Scenes. The dataset consists of synthetic spatial audio mixtures of sound events spatialized for two different spatial formats using real measured room impulse responses (RIRs) measured in various spaces of Tampere University (TAU). The mixtures are generated using the same process as the one used to generate the recordings of the TAU-NIGENS Spatial Sound Scenes 2021 dataset for the DCASE2021 Challenge Task 3. The SELD task setup in DCASE2022 is based on spatial recordings of real scenes, captured in the STARS22 dataset. Since the task setup allows use of external data, these synthetic mixtures serve as additional training material for the baseline model, and they are shared for reasons of reproducibility. For more details on the task setup, please refer to the task description. Note that the generator code and the collection of room responses used to spatialize sound samples will be also be made available soon. For more details on the recording of RIRs, spatialization, and generation, see: Archontis Politis, Sharath Adavanne, Daniel Krause, Antoine Deleforge, Prerak Srivastava, Tuomas Virtanen (2021). A Dataset of Dynamic Reverberant Sound Scenes with Directional Interferers for Sound Event Localization and Detection. In Proceedings of the Detection and Classification of Acoustic Scenes and Events 2020 Workshop (DCASE2021), Barcelona, Spain. available here. SPECIFICATIONS: 13 target sound classes (see task description for details) The sound event samples are sources from the FSD50K dataset, based on affinity of the labels in that dataset to the target classes. The selection on distinguishing which labels in FSD50K corresponded to the target ones, then selecting samples that were tagged with only those labels, and additionally that they had annotator rating of Present and Predominant (see FSD50K for more details). The list of the selected files is included here. 1200 1-minute long spatial recordings Sampling rate of 24kHz Two 4-channel recording formats, first-order Ambisonics (FOA) and tetrahedral microphone array (MIC) Spatial events spatialized in 9 unique rooms, using measured RIRs for the two formats Maximum polyphony of 2 (with possible same-class events overlapping) Even though the whole set is used for training of the baseline without distinction between the mixtures, we have included a separation into a training and testing split, in case on one needs to test the performance purely on those synthetic conditions (for example for comparisons with training on mixed synthetic-real data, fine-tuning on real data, or training on real data only). The training split is indicated as fold1 in the dataset, contains 900 recordings spatialized on 6 rooms (150 recordings/room) and it is based on samples from the development set of FSD50K. The testing split is indicated as fold2 in the dataset, contains 300 recordings spatialized on 3 rooms (100 recordings/room) and it is based on samples from the evaluation set of FSD50K. Common metadata files for both formats are provided. For the file naming and the metadata format, refer to the task setup. FSD50K SELECTION: The list of selected sound event recordings is included along the recordings and metadata, as FSD50K_selected.txt. Each line in the text file has the following structure: [target_label]/[train/test]/[FSD50K_label]/filename.wav with an example: domesticSounds/train/Boiling/16584.wav meaning that the file 16584.wav from FSD50K, with the Boiling label of FSD50K, is included in the samples for the training split of those synthetic recordings, and it is mapped to the target class of domestic sounds. Note that there can be multiple FSD50K labels mapped the same target class. Also note that if these are downloaded from FSD50K, and a folder structure is created that replicates the structure in the list, the resulting folder can be used out-of-the-box with the scene generator to generate new mixtures with the same or different parameters. Note that no sounds form FSD50K have been selected for the Music target class. Background and pop music tracks from the public domain have been cropped and used instead. DOWNLOAD INSTRUCTIONS: Download the zip files and use your preferred compression tool to unzip these split zip files. To extract a split zip archive (named as zip, z01, z02, ...), you could use, for example, the following syntax in Linux or OSX terminal: Combine the split archive to a single archive: zip -s 0 split.zip --out single.zip Extract the single archive using unzip: unzip single.zip数据集说明：本音频数据集为DCASE2022挑战赛任务3：真实空间声景下的声音事件定位与检测的补充材料。本数据集包含采用两种不同空间格式生成的合成空间音频混合片段，其空间化过程使用了在坦佩雷大学（Tampere University, TAU）多个场地采集的实测房间冲激响应（room impulse responses, RIRs）。该混合片段的生成流程与DCASE2021挑战赛任务3所用的TAU-NIGENS 2021空间声景数据集的录制生成流程完全一致。DCASE2022的声音事件定位与检测（SELD）任务设置基于STARS22数据集中采集的真实场景空间录音。由于本次任务设置允许使用外部数据，本合成混合片段可作为基准模型的额外训练素材，且为保证实验可复现性而公开共享。如需了解任务设置的更多细节，请参阅任务说明文档。请注意：用于声音样本空间化的生成代码与房间冲激响应集即将公开。如需了解房间冲激响应采集、空间化处理与生成流程的更多细节，请参阅：Archontis Politis、Sharath Adavanne、Daniel Krause、Antoine Deleforge、Prerak Srivastava、Tuomas Virtanen（2021）。《面向声音事件定位与检测的带定向干扰源的动态混响声景数据集》。收录于2020年声学场景检测与分类研讨会（DCASE2021）论文集，西班牙巴塞罗那。原文链接见此处。 规格参数：13个目标声音类别（详细类别清单请参阅任务说明）。声音事件样本取自FSD50K数据集，选取依据为该数据集的标签与本次任务目标类别的匹配度。具体选取流程为：首先甄别FSD50K数据集中与目标类别对应的标签，随后筛选仅标注有此类标签的样本，且额外要求样本的标注员评分需为「存在且主导」（更多细节请参阅FSD50K数据集文档）。所选文件的清单已附于此处。共1200条1分钟长的空间录音，采样率为24kHz。支持两种4通道录音格式：一阶Ambisonics（first-order Ambisonics, FOA）与四面体麦克风阵列（tetrahedral microphone array, MIC）。空间事件基于两种格式的实测房间冲激响应，在9个独立房间中完成空间化处理。最大复音数为2（允许同类事件重叠）。尽管本数据集整体可直接用于基准模型训练，无需区分混合片段类型，但我们仍将数据集划分为训练集与测试集，以供研究者仅在合成场景条件下测试模型性能（例如用于对比合成-真实混合数据训练、基于真实数据微调、仅使用真实数据训练等实验场景）。训练集在数据集中以折1（fold1）标识，包含900条录音，分别来自6个空间化房间（每个房间150条录音），其样本取自FSD50K数据集的开发集。测试集在数据集中以折2（fold2）标识，包含300条录音，分别来自3个空间化房间（每个房间100条录音），其样本取自FSD50K数据集的评估集。两种格式共用元数据文件。关于文件命名规则与元数据格式，请参阅任务设置文档。 FSD50K数据集选取说明：所选声音事件录音的清单已随录音与元数据一同提供，文件名为"FSD50K_selected.txt"。文本文件中每行的格式为：[目标类别]/[训练集/测试集]/[FSD50K标签]/[文件名.wav]，示例："domesticSounds/train/Boiling/16584.wav"，其含义为：来自FSD50K的16584.wav文件，其FSD50K标签为Boiling，被纳入本次合成录音的训练集样本，且映射至家庭声音这一目标类别。请注意：多个FSD50K标签可映射至同一目标类别。同时请注意，若从FSD50K下载对应样本，并按照清单中的目录结构创建文件夹，则生成的文件夹可直接与声景生成工具配合，用于生成参数相同或不同的新混合片段。请注意：音乐目标类别未选取任何FSD50K数据集的声音样本，取而代之的是从公有领域裁剪的背景音与流行音乐曲目。 下载说明：下载分卷压缩包后，使用您偏好的解压工具进行解压。若需在Linux或macOS终端中解压分卷压缩归档文件（命名格式为zip、z01、z02……），可参考以下命令：首先将分卷归档合并为单个归档文件：zip -s 0 split.zip --out single.zip 随后使用unzip命令解压单个归档文件：unzip single.zip

搜集汇总

背景与挑战

背景概述

该数据集是DCASE2022挑战赛任务3的合成训练数据集，用于声音事件定位与检测（SELD）的基线模型训练。它包含1200条1分钟长的空间音频混合，基于FSD50K数据集的声音样本和真实测量的房间脉冲响应（RIRs）生成，支持两种空间格式（一阶Ambisonics和四面体麦克风阵列），并分为训练集和测试集以促进模型评估。