Skip to main content
EURASIP Journal on Audio, Speech, and Music Processing

Table 1 Dataset descriptions

From: Deep neural network-based bottleneck feature and denoising autoencoder-based dereverberation for distant-talking speaker identification

Data type

Usage

Data set

Training data

To train the DNN and GMM

100 (speakers) × 5 (utterances) × 3 (environments)

Development data

To determine the settings of the DNN (layers, batches, etc.) in the experimental step

Same as above

Test data

To test the speakers in this dataset in the evaluation step

100 (speakers) × 20 (utterances) × 5 (environments)

Back to article page