Table 1.
Breakdown of the EMIR dataset.
Fig 1.
EKM architecture used in experiments.
Fig 2.
Experiment 1: EKM confusion matrices using FilterBank on EMIR.
Fig 3.
Experiment 1: EKM confusion matrices using MelSpec on EMIR.
Fig 4.
Experiment 1: EKM confusion matrices using Chroma on EMIR.
Fig 5.
Experiment 1: EKM confusion matrices using MFCC on EMIR.
Table 2.
Experiment 1: Recognition accuracies of VGG networks on EMIR using FilterBank, MelSpec, MFCC and Chroma features with 3s samples.
Table 3.
Experiment 2: EKM model accuracy using 1s, 3s and 5s sample lengths, MFCC features and EMIR data.
Fig 6.
Experiment 2: Convergence curve in 250-epoch training.
Table 4.
Experiment 3: Comparison of EKM with other CNN and LSTM models, all applied to the EMIR dataset.
Table 5.
Statistical significance tests for Experiment 3.