Paula D. P. Costa

Lucas H. Ueda, João G. T. Lima, Paula D. P. Costa

Speech Emotion Recognition (SER) in real-world scenarios remains challenging due to severe class imbalance and the prevalence of spontaneous, natural speech. While recent approaches leverage self-supervised learning (SSL) representations and multimodal fusion of speech and text, most existing methods apply supervision only at the final classification layer, limiting the discriminative power of intermediate representations. In this work, we propose Crab (Contrastive Representation and Multimodal Aligned Bottleneck), a bimodal Cross-Modal Transformer architecture that integrates speech representations from WavLM and textual representations from RoBERTa, together with a novel \textit{Multi Layer Contrastive Supervision} (MLCS) strategy. MLCS injects multi-positive contrastive learning signals at multiple layers of the network, encouraging emotionally discriminative representations throughout the model without introducing additional parameters at inference time. To further address data imbalance, we adopt weighted cross-entropy during training. We evaluate the proposed approach on three benchmark datasets covering different degrees of emotional naturalness: IEMOCAP, MELD, and MSP-Podcast 2.0. Experimental results demonstrate that Crab consistently outperforms strong unimodal and multimodal baselines across all datasets, with particularly large gains under naturalistic and highly imbalanced conditions. These findings highlight the effectiveness of \textit{Multi Layer Contrastive Supervision} as a general and robust strategy for SER. Official implementation can be found in https://github.com/AI-Unicamp/Crab.

Gabriel Dias, Rodrigo Pommot Berto, Mateus Oliveira et al.

Purpose: To investigate the use of a Vision Transformer (ViT) to reconstruct/denoise GABA-edited magnetic resonance spectroscopy (MRS) from a quarter of the typically acquired number of transients using spectrograms. Theory and Methods: A quarter of the typically acquired number of transients collected in GABA-edited MRS scans are pre-processed and converted to a spectrogram image representation using the Short-Time Fourier Transform (STFT). The image representation of the data allows the adaptation of a pre-trained ViT for reconstructing GABA-edited MRS spectra (Spectro-ViT). The Spectro-ViT is fine-tuned and then tested using \textit{in vivo} GABA-edited MRS data. The Spectro-ViT performance is compared against other models in the literature using spectral quality metrics and estimated metabolite concentration values. Results: The Spectro-ViT model significantly outperformed all other models in four out of five quantitative metrics (mean squared error, shape score, GABA+/water fit error, and full width at half maximum). The metabolite concentrations estimated (GABA+/water, GABA+/Cr, and Glx/water) were consistent with the metabolite concentrations estimated using typical GABA-edited MRS scans reconstructed with the full amount of typically collected transients. Conclusion: The proposed Spectro-ViT model achieved state-of-the-art results in reconstructing GABA-edited MRS, and the results indicate these scans could be up to four times faster.

Lucas H. Ueda, João G. T. Lima, Pedro R. Corrêa et al.

This paper presents SelfTTS, a text-to-speech (TTS) model designed for cross-speaker style transfer that eliminates the need for external pre-trained speaker or emotion encoders. The architecture achieves emotional expressivity in neutral speakers through an explicit disentanglement strategy utilizing Gradient Reversal Layers (GRL) combined with cosine similarity loss to decouple speaker and emotion information. We introduce Multi Positive Contrastive Learning (MPCL) to induce clustered representations of speaker and emotion embeddings based on their respective labels. Furthermore, SelfTTS employs a self-refinement strategy via Self-Augmentation, exploiting the model's voice conversion capabilities to enhance the naturalness of synthesized speech. Experimental results demonstrate that SelfTTS achieves superior emotional naturalness (eMOS) and robust stability in target timbre and emotion compared to state-of-the-art baselines.