VOICe

声音由三种不同声音事件 (“婴儿哭泣”，“玻璃破碎” 和 “枪声”) 的1449不同混合组成:
三种不同类别的声学场景 (“车辆” 、 “室外” 和 “室内”) 的具有背景噪声的1242混合物，混合在2 SNR值 (-3，-9 dB) 下，即207混合物x 3声学场景x 2 SNRs = 1242
没有任何背景噪音的207混合物。

This dataset contains 1449 distinct audio mixtures formed from three types of sound events: "baby crying", "glass breaking", and "gunshot".
Among these, 1242 mixtures are blended with background noise from three acoustic scene categories: "vehicle", "outdoor", and "indoor", at two signal-to-noise ratio (SNR) levels (-3 dB and -9 dB). The total number of these noise-containing mixtures is calculated as 207 mixtures × 3 acoustic scenarios × 2 SNRs = 1242.
There are an additional 207 mixtures without any background noise.

搜集汇总

背景与挑战

背景概述

VOICe是一个用于声音事件检测的音频数据集，包含三种特定声音事件在不同声学场景和信噪比条件下的混合样本，总样本量为1449个。该数据集专为领域自适应研究设计，具有明确的实验控制和场景多样性。