ASMR_Dataset
收藏资源简介:
该数据集是一个经过手动清理和自动处理的多媒体数据集,包括视频字幕和对应的音频。数据集中的字幕经过筛选,确保包含中文字幕,并且时间戳正确对齐。通过使用mel band roformer模型,数据集中的音频去除了背景杂音,如SE(特效声)、BGM(背景音乐)、口水声等。数据集还通过分离左右声道并识别出最响亮的声道来进一步处理音频数据,最终创建了索引以便于使用。
This dataset is a manually cleaned and automatically processed multimedia dataset containing video subtitles and their corresponding audio. The subtitles in the dataset have been filtered to ensure they include Chinese subtitles with correctly aligned timestamps. Utilizing the mel band RoFormer model, the audio in the dataset has been denoised to remove background noises such as sound effects (SE), background music (BGM), and saliva sounds. The dataset further processes the audio data by separating the left and right audio channels and identifying the loudest channel, and finally creates indexes for easy access and utilization.




