Korean Emergency Audio and Speech dataset

该数据集包含16种紧急情况，总时长约3500小时的音频资料，涵盖了求助呼叫、火灾、气体事故和暴力犯罪等多种紧急场景。此外，该数据集特别关注六种求助呼叫场景，在训练集和测试集中分别提供了186,000个和18,000个样本。在规模上，训练集有422,000个样本，而测试集有55,000个样本。该数据集的任务是进行求助呼叫检测。

This dataset contains audio data with a total duration of approximately 3,500 hours, covering 16 categories of emergency scenarios including distress calls, fires, gas incidents, violent crimes and other common emergency situations. Specifically, this dataset places special emphasis on six types of distress call scenarios, with 186,000 and 18,000 samples allocated to its training set and test set respectively. In terms of overall scale, the training set consists of 422,000 samples while the test set contains 55,000 samples. The core task of this dataset is distress call detection.