JST-SUPERB/MUSAN-speech_unit_part2

该数据集包含多个音频编码格式和采样率的分割，每个分割包含语音输入、不同信噪比下的噪声音频及其转录文本，以及干净音频的转录文本和单位序列。数据集的总大小为27616875746.349995字节，下载大小为13654553512字节。

This dataset contains multiple splits with different audio encoding formats and sampling rates. Each split includes speech input, noisy audio at various signal-to-noise ratios and their transcriptions, as well as transcriptions and unit sequences of clean audio. The total size of the dataset is 27616875746.349995 bytes, and the download size is 13654553512 bytes.

原始信息汇总

数据集概述

数据集名称

MUSAN-speech_unit_part2

数据集配置

• 配置名称: default
• 数据文件:
  • split: encodec_24k_12bps
    • path: data/encodec_24k_12bps-*
  • split: encodec_24k_1_5bps
    • path: data/encodec_24k_1_5bps-*
  • split: encodec_24k_24bps
    • path: data/encodec_24k_24bps-*
  • split: encodec_24k_3bps
    • path: data/encodec_24k_3bps-*
  • split: encodec_24k_6bps
    • path: data/encodec_24k_6bps-*
  • split: funcodec_en_libritts_16k_gr1nq32ds320
    • path: data/funcodec_en_libritts_16k_gr1nq32ds320-*
  • split: funcodec_en_libritts_16k_gr8nq32ds320
    • path: data/funcodec_en_libritts_16k_gr8nq32ds320-*
  • split: funcodec_en_libritts_16k_nq32ds320
    • path: data/funcodec_en_libritts_16k_nq32ds320-*
  • split: funcodec_en_libritts_16k_nq32ds640
    • path: data/funcodec_en_libritts_16k_nq32ds640-*
  • split: funcodec_zh_en_16k_nq32ds320
    • path: data/funcodec_zh_en_16k_nq32ds320-*
  • split: funcodec_zh_en_16k_nq32ds640
    • path: data/funcodec_zh_en_16k_nq32ds640-*

数据集信息

• 特征:
  • name: speech_input
    • dtype: string
  • name: noisy_-20dB
    • dtype: audio
  • name: noisy_10dB_transcription_whisper-small.en
    • dtype: string
  • name: noisy_5dB_transcription_whisper-small.en
    • dtype: string
  • name: noisy_0dB_transcription_whisper-small.en
    • dtype: string
  • name: noisy_-5dB_transcription_whisper-small.en
    • dtype: string
  • name: noisy_-10dB_transcription_whisper-small.en
    • dtype: string
  • name: noisy_10dB_transcription_whisper-medium.en
    • dtype: string
  • name: noisy_5dB_transcription_whisper-medium.en
    • dtype: string
  • name: noisy_0dB_transcription_whisper-medium.en
    • dtype: string
  • name: noisy_-5dB_transcription_whisper-medium.en
    • dtype: string
  • name: noisy_-10dB_transcription_whisper-medium.en
    • dtype: string
  • name: noisy_10dB_transcription_whisper-large-v3
    • dtype: string
  • name: noisy_5dB_transcription_whisper-large-v3
    • dtype: string
  • name: noisy_0dB_transcription_whisper-large-v3
    • dtype: string
  • name: noisy_-5dB_transcription_whisper-large-v3
    • dtype: string
  • name: noisy_-10dB_transcription_whisper-large-v3
    • dtype: string
  • name: output
    • dtype: string
  • name: clean_audio_transcription_whisper-small.en
    • dtype: string
  • name: clean_audio_transcription_whisper-medium.en
    • dtype: string
  • name: clean_audio_transcription_whisper-large-v3
    • dtype: string
  • name: clean_audio_unit
    • sequence: int64
  • name: noisy_10dB_unit
    • sequence: int64
  • name: noisy_5dB_unit
    • sequence: int64
  • name: noisy_0dB_unit
    • sequence: int64
  • name: noisy_-5dB_unit
    • sequence: int64
  • name: noisy_-10dB_unit
    • sequence: int64

数据集分割

• name: encodec_24k_12bps
  • num_bytes: 2569154395.85
  • num_examples: 5135
• name: encodec_24k_1_5bps
  • num_bytes: 1300320619.85
  • num_examples: 5135
• name: encodec_24k_24bps
  • num_bytes: 4019250139.85
  • num_examples: 5135
• name: encodec_24k_3bps
  • num_bytes: 1481582587.85
  • num_examples: 5135
• name: encodec_24k_6bps
  • num_bytes: 1844106523.85
  • num_examples: 5135
• name: funcodec_en_libritts_16k_gr1nq32ds320
  • num_bytes: 3055196635.85
  • num_examples: 5135
• name: funcodec_en_libritts_16k_gr8nq32ds320
  • num_bytes: 3055196635.85
  • num_examples: 5135
• name: funcodec_en_libritts_16k_nq32ds320
  • num_bytes: 3055052251.85
  • num_examples: 5135
• name: funcodec_en_libritts_16k_nq32ds640
  • num_bytes: 2090981851.85
  • num_examples: 5135
• name: funcodec_zh_en_16k_nq32ds320
  • num_bytes: 3055052251.85
  • num_examples: 5135
• name: funcodec_zh_en_16k_nq32ds640
  • num_bytes: 2090981851.85
  • num_examples: 5135

数据集大小

• download_size: 13654553512
• dataset_size: 27616875746.349995