EML-CNN reports 96%+ speaker-identification accuracy across three speech benchmarks
An EML-CNN combining dense convolutional blocks, gated recurrent units and attention achieved speaker-identification accuracies above 96% across three benchmark datasets.
