HIGHLIGHTS
SUMMARY
As the authors will be further explored from Section V onward, when used as a feature extractor that feeds into quaternion neural_networks (QNNs), RH-emo improves the performance in SER tasks while considerably reducing the number of trainable parameters and computing resources, compared to equivalent real-valued models processing plain spectrograms. The authors define a novel method, RH-emo, that draws quaternion-valued embeddings from speech signals, where each quaternion component is tailored to a specific emotional characteristic.. The authors leverage the capabilities of quaternion emotion embeddings and the effectiveness of quaternion convolutional neural_networks . . .
If you want to have access to all the content you need to log in!
Thanks :)
If you don't have an account, you can create one here.