TIMIT, Aurora-4, AMI, and LibriSpeech

Four different corpora are used for our experiments, which are TIMIT, Aurora-4, AMI, and LibriSpeech. TIMIT contains broadband 16kHz recordings of phonetically-balanced read speech. Aurora-4 is another broadband corpus designed for noisy speech recognition tasks, based on the Wall Street Journal corpus (WSJ0). AMI corpus consists of 100 hours of meeting recordings, recorded in three different meeting rooms with different acoustic properties, and with multiple microphones. LibriSpeech corpus contains 1,000 hours of read speech sampled at 16kHz.