LibriSpeech is a corpus of approximately 1000 hours of read English speech with sampling rate of 16 kHz,
prepared by Vassil Panayotov with the assistance of Daniel Povey. The data is derived from read
audiobooks from the LibriVox project, and has been carefully segmented and aligned.87
