地震、爆炸、海难、海啸、火灾、空难、矿难、战争八个场景语音识别音频合集

地震、爆炸、海难、海啸、火灾、空难、矿难、战争八个场景语音识别音频合集，内含1个数据集说明文件，8个音频合集数据库。采集方案：模拟不同场景的环境噪声，并通过不同人员在该回放场景下进行现场录音的方式收录。在普通安静实验室、卧室等通过苹果手机登录拾音APP进行高保真数据采集，人员需要以惊恐、平静、虚弱三个状态进行模拟；加工方式：F8n录音机+4麦在不同场景下回流长数据，长数据爱标客平台上标注，生成TextGrid文件，通过vad切割工具进行长数据切分短句子。数据集用于完成地震环境下语音智能识别能力的验证测试。This is a collection of speech recognition audio datasets covering 8 scenarios: earthquake, explosion, shipwreck, tsunami, fire, air crash, mining accident, and war. It contains 1 dataset description file and 8 audio collection databases. Collection scheme: First, simulate ambient noise corresponding to each scenario, then have different personnel conduct on-site recording in the playback scenarios with the simulated noise. Additionally, perform high-fidelity data collection via an audio recording app logged in on an Apple iPhone in ordinary quiet environments such as laboratories and bedrooms, where the personnel are required to simulate three states: panic, calm, and weakness. Processing method: Record long-duration audio data in various scenarios using an F8n recorder equipped with 4 microphones. Annotate the long audio data on the AibiaoKe platform, generate TextGrid files, and split the long audio data into short sentences using a VAD (Voice Activity Detection) cutting tool. This dataset is intended for the verification and testing of speech intelligent recognition capabilities in earthquake environments.

搜集汇总

背景与挑战

背景概述

该数据集是一个语音识别音频合集，涵盖地震、爆炸、海难、海啸、火灾、空难、矿难和战争八个典型灾害场景。它通过模拟不同场景的环境噪声和人员状态进行采集，并经过专业加工处理，旨在用于地震环境下语音智能识别能力的验证测试。