Simulated Dataset for TD-GWF Evaluation

该数据集包含了20,000个训练样本、5,000个验证样本以及3,000个测试样本，每个样本时长为4秒，采样率为16kHz。数据集中包含两种语音信号和一种噪声信号，它们之间的重叠比例和信噪比水平各不相同。此外，该数据集是通过模拟的房间脉冲响应生成的，并融入了多种麦克风配置。其规模为20,000个训练样本、5,000个验证样本以及3,000个测试样本，所面临的任务是语音分离和波束形成。

This dataset consists of 20,000 training samples, 5,000 validation samples, and 3,000 test samples. Each sample has a duration of 4 seconds and a sampling rate of 16 kHz. The dataset contains two categories of speech signals and one category of noise signal, with varying overlap ratios and signal-to-noise ratio (SNR) levels across these signals. Additionally, this dataset is generated via simulated room impulse responses and integrates multiple microphone configurations. The core tasks for this dataset are speech separation and beamforming.