Search papers, labs, and topics across Lattice.
2
0
3
2
Tri-view Consensus-Guided Fusion (TriCGF) is proposed for jointly modeling spectrogram, Mel-frequency cepstral coefficients, and HuBERT representations, outperforming representative SER methods on both datasets.
Fine-tuning not only enhances the decoder's access to existing audio token evidence but also reveals that crucial event information is already present before any adjustments are made.
