SPARCO-project/benchmark_DEMAND_noise

该数据集是从DEMAND（多样环境多通道声学噪声数据库）中分割出来的子集，专门为SPARCO噪声消融基准测试准备。其用途是提供固定的4秒环境噪声片段，用于基于AUROC的SAE噪声相关特征选择、二元噪声存在评分器训练、评分器阈值校准以及最终保留基准评估。数据集采用环境级分割以减少泄漏风险，包含三个分割：feature_scorer_train（用于特征选择和评分器训练）、scorer_val（用于阈值校准）和benchmark_test（仅用于最终评估）。原始DEMAND数据经过处理，仅使用16 kHz文件夹和通道1，分割成非重叠的4秒片段，并保存为16位PCM WAV文件，未应用去噪、增强、归一化或人工混合处理。This dataset is a segmented subset derived from DEMAND: Diverse Environments Multichannel Acoustic Noise Database. It is prepared for the SPARCO noise ablation benchmark. The intended use is to provide fixed 4-second environmental noise segments for AUROC-based SAE noise-related feature selection, binary noise-presence scorer training, scorer threshold calibration, and final held-out benchmark evaluation. The dataset uses environment-level splitting to reduce leakage risk, with three splits: feature_scorer_train (for feature selection and scorer training), scorer_val (for threshold calibration), and benchmark_test (for final evaluation only). Original DEMAND recordings were processed by using only 16 kHz folders and channel 1, segmenting into 4-second non-overlapping clips, and saving as 16-bit PCM WAV files, with no denoising, enhancement, normalization, or artificial mixing applied.