radiotalk-us-audio-grok-noisy
收藏资源简介:
radiotalk-us-audio-grok-noisy 是一个用于空中交通管制(ATC)领域的合成语音数据集,专注于VHF-AM通信信道的模拟降级场景。该数据集是 twangodev/radiotalk-us-audio-grok-clean 的噪声版本,每个干净的语音样本包含三个独立的降级变体,包括带通滤波、噪声、衰落、外差、PTT点击、编解码器伪影等,模拟了真实无线电通信中的多种失真。当前版本包含791,703条音频样本,来自 twangodev/radiotalk-us-transcripts-grok-4.20-50k 的前25,000个转录文本,并计划逐步扩展到全部50,000个转录文本。数据集适用于自动语音识别(ASR)模型的训练与评估,尤其是针对航空通信中噪声环境下的语音识别任务。许可证为CC-BY-4.0,需要标注radiotalk项目及xAI提供的底层TTS。
radiotalk-us-audio-grok-noisy is a synthetic speech dataset for the air traffic control (ATC) domain, focusing on simulated degradation of VHF-AM communication channels. It is a noisy version of twangodev/radiotalk-us-audio-grok-clean, with each clean speech sample containing three independent degraded variants including bandpass filtering, noise, fading, heterodyne, PTT clicks, codec artifacts, etc., simulating various distortions in real radio communications. The current version contains 791,703 audio samples derived from the first 25,000 transcripts of twangodev/radiotalk-us-transcripts-grok-4.20-50k, with plans to gradually extend to all 50,000 transcripts. The dataset is suitable for training and evaluating automatic speech recognition (ASR) models, especially for speech recognition in noisy environments in aviation communications. Licensed under CC-BY-4.0, with attribution required to the radiotalk project and the underlying TTS provided by xAI.
radiotalk-us-audio-grok-noisy 数据集概述
基本信息
- 数据集名称:radiotalk-us-audio-grok-noisy(Grok TTS生成的美国航空无线电通话语音数据,含噪声版本)
- 许可证:CC-BY-4.0
- 任务类型:自动语音识别(Automatic Speech Recognition)
- 语言:英语
- 数据集规模:100K < n < 1M
- 当前数据量:791,703 行
数据集内容
- 数据来源:基于
twangodev/radiotalk-us-transcripts-grok-4.20-50k数据集中的前25,000条转录文本生成,后续将分阶段扩展至完整的50,000条。 - 数据性质:合成语音数据,由Grok TTS生成,覆盖航空管制(ATC)和航空领域的VHF无线电通话内容。
- 噪声处理:每一条干净语音均提供3个独立退化变体,模拟VHF-AM信道退化效果,具体退化类型包括:
- 带通滤波(bandpass)
- 噪声叠加(noise)
- 信号衰落(fading)
- 差拍干扰(heterodyne)
- PTT按键声(PTT clicks)
- 编解码器伪影(codec artifacts)
- 退化管线:与higgs/tada噪声数据集使用相同的radiotalk无线电管线。
数据集用途
- 主要用于自动语音识别任务的训练与评估,特别适用于航空无线电通信场景下的抗噪语音识别研究。
引用要求
- 使用该数据集时需标注radiotalk项目及xAI(作为底层TTS提供方)。




