Recreating Neural Activity During Speech Production with Language and Speech Model Embeddings Khanday, Owais Mujtaba Rodríguez San Esteban, Pablo Ahmad, Zubair Ouellet, Marc González López, José Andrés Understanding how neural activity encodes speech and language production is a fundamental challenge in neuroscience and artificial intelligence. This study investigates whether embeddings from large-scale, self-supervised language and speech models can effectively reconstruct high-gamma neural activity characteristics, key indicators of cortical processing, recorded during speech production. We use pre-trained embeddings from deep learning models on linguistic and acoustic data to map high-level speech features onto high-gamma signals. We analyze the extent to which these embeddings preserve the spatio-temporal dynamics of brain activity. Reconstructed neural signals are evaluated against high-gamma ground-truth activity using correlation metrics and signal reconstruction quality assessments. The results indicate High-gamma activity was effectively reconstructed using language and speech model embeddings, yielding Pearson correlation coefficients of 0.79–0.99 across all participants. 2025-09-02T07:09:47Z 2025-09-02T07:09:47Z 2025-08-17 conference output Khanday, O.M., Esteban, P.R.S., Lone, Z.A., Ouellet, M., Gonzalez-Lopez, J.A. (2025) Recreating Neural Activity During Speech Production with Language and Speech Model Embeddings. Proc. Interspeech 2025, 5553-5557, doi: 10.21437/Interspeech.2025-1400 https://hdl.handle.net/10481/105970 10.21437/Interspeech.2025-1400 eng http://creativecommons.org/licenses/by-nc-sa/4.0/ open access Atribución-NoComercial-CompartirIgual 4.0 Internacional ISCA