MultiSV

MultiSV数据集由布尔诺理工大学信息技术学院语音技术研究组创建，旨在为远场多通道语音验证系统提供训练和评估材料。该数据集通过在Voxceleb数据集的干净部分上进行数据模拟来解决多通道训练数据的缺乏问题。数据集内容包括约77小时每麦克风的模拟4麦克风阵列数据，涵盖背景噪声和混响。创建过程中，使用了房间脉冲响应生成技术。MultiSV数据集适用于语音增强、去噪和语音清晰度提升等模型的训练，特别适用于解决远场环境下的语音识别问题。

The MultiSV dataset was developed by the Speech Technology Research Group at the Faculty of Information Technology, Brno University of Technology, with the goal of providing training and evaluation materials for far-field multi-channel speech verification systems. To address the scarcity of multi-channel training data, this dataset is constructed through data simulation on the clean subset of the Voxceleb dataset. It contains approximately 77 hours of simulated 4-microphone array data per microphone, incorporating background noise and reverberation. Room impulse response (RIR) generation techniques were utilized during the dataset's creation. The MultiSV dataset is applicable for training models including speech enhancement, denoising, and speech clarity enhancement, and is particularly suited for solving speech recognition problems in far-field environments.