Cnam-LMSSC/french_librispeech_vibravoxed

该数据集是facebook/multilingual_librispeech法语部分的一个版本，名为Vibravoxed。数据集包含多种音频特征，如airborne、transcript、speaker_id以及多种模拟的音频数据（如throat_microphone_simulated、rigid_in_ear_microphone_simulated等），所有音频数据的采样率均为16000 Hz。数据集分为train、dev、test、1_hours和9_hours等多个部分，每个部分都有对应的字节大小和示例数量。此外，该数据集使用了反向EBEN模型对干净的语音进行了恶化处理。

This dataset includes multiple audio features such as airborne, throat_microphone_simulated, etc., each with a sampling rate of 16000. Additionally, it contains transcriptions and speaker IDs. The dataset is divided into several parts including train, dev, test sets, and subsets with specific durations. The download size and total size of the dataset are also provided. The dataset is based on the French split of facebook/multilingual_librispeech and has been deteriorated using specific reverse EBEN models.