@speechpapers.bsky.social

@speechpapers.bsky.social

362 Followers 305 Following

Every day I post short descriptions (well titles, really) of recent speech articles

The flowchart outlines the speech emotion recognition process, starting with four datasets: CREMA-D, RAVDESS, EMO-DB, and ASVP-ESD. It details preprocessing steps, including silence removal, resampling, spectral gating, and min-max scaling. Feature extraction employs Continuous Wavelet Transform (CWT) and dual attention mechanisms (Spatial and Channel Attention) to derive emotion-relevant features. The classification uses DUST-Net, incorporating deformable convolution, Hierarchical Swin Transformer, and token pruning with Uncertainty-Aware Attention, leading to emotion outputs for Happy, Sad, Surprise, Disgust, Fear, Neutral, Angry, and Bored.
Various English translations of the first few lines of The Odyssey
A silly picture to celebrate 300 followers that rifts off the film 300 but has an academic holding books rather than a muscular Spartan with a spear
what https://www.whatsmydns.net/#A/pubmed.ncbi.nlm.nih.gov shows