The field of speech and brain-computer interfaces is rapidly advancing with innovative solutions for ultra-low bitrate speech compression, brain-guided image synthesis, and neural-driven avatar synthesis. Researchers are exploring new approaches to decode brain signals into speech, gestures, and facial expressions, enabling more natural and intuitive human-computer interaction. Noteworthy papers include STCTS, which achieves a 75x bitrate reduction for speech compression, and NeuroVolve, which generates coherent scenes that satisfy complex neural objectives. Additionally, Mind-to-Face decodes EEG signals into photorealistic facial expressions, while Large Speech Model enabled Semantic Communication enables adaptive transmission over lossy channels. These breakthroughs have the potential to revolutionize the way we interact with technology and each other.
Breakthroughs in Speech and Brain-Computer Interfaces
Sources
STCTS: Generative Semantic Compression for Ultra-Low Bitrate Speech via Explicit Text-Prosody-Timbre Decomposition
NeuroVolve: Evolving Visual Stimuli toward Programmable Neural Objectives
fMRI2GES: Co-speech Gesture Reconstruction from fMRI Signal with Dual Brain Decoding Alignment
MEGConformer: Conformer-Based MEG Decoder for Robust Speech and Phoneme Classification
VibOmni: Towards Scalable Bone-conduction Speech Enhancement on Earables
Mind-to-Face: Neural-Driven Photorealistic Avatar Synthesis via EEG Decoding
Neural Decoding of Overt Speech from ECoG Using Vision Transformers and Contrastive Representation Learning
Large Speech Model Enabled Semantic Communication