MEDUSA: A Multimodal Deep Fusion Multi-Stage Training Framework for Speech Emotion Recognition in Naturalistic Conditions

• URL: http://arxiv.org/abs/2506.09556v1
• Date: Wed, 11 Jun 2025 09:41:23 GMT
• Title: MEDUSA: A Multimodal Deep Fusion Multi-Stage Training Framework for Speech Emotion Recognition in Naturalistic Conditions
• Authors: Georgios Chatzichristodoulou, Despoina Kosmopoulou, Antonios Kritikos, Anastasia Poulopoulou, Efthymios Georgiou, Athanasios Katsamanis, Vassilis Katsouros, Alexandros Potamianos,
• Abstract summary: MEDUSA is a multimodal framework with a four-stage training pipeline.<n>DeepSER is a novel extension of a deep cross-modal transformer fusion mechanism.<n>Manor MixUp is employed for further regularization.
• Score: 46.34220791244788
• License: http://creativecommons.org/licenses/by-sa/4.0/
• Abstract: SER is a challenging task due to the subjective nature of human emotions and their uneven representation under naturalistic conditions. We propose MEDUSA, a multimodal framework with a four-stage training pipeline, which effectively handles class imbalance and emotion ambiguity. The first two stages train an ensemble of classifiers that utilize DeepSER, a novel extension of a deep cross-modal transformer fusion mechanism from pretrained self-supervised acoustic and linguistic representations. Manifold MixUp is employed for further regularization. The last two stages optimize a trainable meta-classifier that combines the ensemble predictions. Our training approach incorporates human annotation scores as soft targets, coupled with balanced data sampling and multitask learning. MEDUSA ranked 1st in Task 1: Categorical Emotion Recognition in the Interspeech 2025: Speech Emotion Recognition in Naturalistic Conditions Challenge.

Related papers

• Emotion-LLaMAv2 and MMEVerse: A New Framework and Benchmark for Multimodal Emotion Understanding [45.13650362585136]
  We present Emotion-LLaMAv2 and the MMEVerse benchmark, establishing an end-to-end pipeline together with a standardized evaluation setting for emotion recognition and reasoning.<n>An end-to-end multiview encoder eliminates external face detection and captures nuanced emotional cues via richer spatial and temporal multiview tokens.<n>A perception-to-cognition curriculum instruction tuning scheme within the LLaMA2 backbone unifies emotion recognition and free-form emotion reasoning.
  arXiv  Detail & Related papers  (2026-01-23T05:02:43Z)
• A Unified Spoken Language Model with Injected Emotional-Attribution Thinking for Human-like Interaction [50.05919688888947]
  This paper presents a unified spoken language model for emotional intelligence, enhanced by a novel data construction strategy termed Injected Emotional-Attribution Thinking (IEAT)<n>IEAT incorporates user emotional states and their underlying causes into the model's internal reasoning process, enabling emotion-aware reasoning to be internalized rather than treated as explicit supervision.<n> Experiments on the Human-like Spoken Dialogue Systems Challenge (HumDial) Emotional Intelligence benchmark demonstrate that the proposed approach achieves top-ranked performance across emotional trajectory modeling, emotional reasoning, and empathetic response generation.
  arXiv  Detail & Related papers  (2026-01-08T14:07:30Z)
• Taming Transformer for Emotion-Controllable Talking Face Generation [61.835295250047196]
  We propose a novel method to tackle the emotion-controllable talking face generation task discretely.<n>Specifically, we employ two pre-training strategies to disentangle audio into independent components and quantize videos into combinations of visual tokens.<n>We conduct experiments on the MEAD dataset that controls the emotion of videos conditioned on multiple emotional audios.
  arXiv  Detail & Related papers  (2025-08-20T02:16:52Z)
• Beyond Classification: Towards Speech Emotion Reasoning with Multitask AudioLLMs [47.325269852330884]
  We develop a strategy to enhance emotion recognition by producing semantically aligned, evidence-grounded explanations.<n>We introduce a unified framework combining reasoning-augmented data supervision, dual-encoder architecture, and task-alternating training.<n> Experiments on IEMOCAP and MELD show that our approach not only improves emotion prediction accuracy but also enhances the coherence and evidential grounding of the generated responses.
  arXiv  Detail & Related papers  (2025-06-07T14:52:58Z)
• OpenOmni: Advancing Open-Source Omnimodal Large Language Models with Progressive Multimodal Alignment and Real-Time Self-Aware Emotional Speech Synthesis [73.03333371375]
  name is a two-stage training framework that integrates omnimodal alignment and speech generation.<n>It surpasses state-of-the-art models across omnimodal, vision-language, and speech-language benchmarks.<n>name achieves real-time speech generation with 1s latency at non-autoregressive mode.
  arXiv  Detail & Related papers  (2025-01-08T15:18:09Z)
• SIFToM: Robust Spoken Instruction Following through Theory of Mind [51.326266354164716]
  We present a cognitively inspired model, Speech Instruction Following through Theory of Mind (SIFToM), to enable robots to pragmatically follow human instructions under diverse speech conditions. Results show that the SIFToM model outperforms state-of-the-art speech and language models, approaching human-level accuracy on challenging speech instruction following tasks.
  arXiv  Detail & Related papers  (2024-09-17T02:36:10Z)
• MSP-Podcast SER Challenge 2024: L'antenne du Ventoux Multimodal Self-Supervised Learning for Speech Emotion Recognition [12.808666808009926]
  We submit to the 2024 edition of the MSP-Podcast Speech Emotion Recognition (SER) Challenge. This challenge is divided into two distinct tasks: Categorical Emotion Recognition and Emotional Attribute Prediction. Our approach employs an ensemble of models, each trained independently and then fused at the score level using a Support Vector Machine (SVM) This joint training methodology aims to enhance the system's ability to accurately classify emotional states.
  arXiv  Detail & Related papers  (2024-07-08T08:52:06Z)
• BLSP-Emo: Towards Empathetic Large Speech-Language Models [34.62210186235263]
  We present BLSP-Emo, a novel approach to developing an end-to-end speech-language model capable of understanding both semantics and emotions in speech. Our experiments demonstrate that the BLSP-Emo model excels in comprehending speech and delivering empathetic responses.
  arXiv  Detail & Related papers  (2024-06-06T09:02:31Z)
• UMETTS: A Unified Framework for Emotional Text-to-Speech Synthesis with Multimodal Prompts [64.02363948840333]
  UMETTS is a novel framework that leverages emotional cues from multiple modalities to generate highly expressive and emotionally resonant speech.<n>EP-Align employs contrastive learning to align emotional features across text, audio, and visual modalities, ensuring a coherent fusion of multimodal information.<n>EMI-TTS integrates the aligned emotional embeddings with state-of-the-art TTS models to synthesize speech that accurately reflects the intended emotions.
  arXiv  Detail & Related papers  (2024-04-29T03:19:39Z)
• Attention-based Interactive Disentangling Network for Instance-level Emotional Voice Conversion [81.1492897350032]
  Emotional Voice Conversion aims to manipulate a speech according to a given emotion while preserving non-emotion components. We propose an Attention-based Interactive diseNtangling Network (AINN) that leverages instance-wise emotional knowledge for voice conversion.
  arXiv  Detail & Related papers  (2023-12-29T08:06:45Z)
• Audio is all in one: speech-driven gesture synthetics using WavLM pre-trained model [2.827070255699381]
  diffmotion-v2 is a speech-conditional diffusion-based generative model with WavLM pre-trained model. It can produce individual and stylized full-body co-speech gestures only using raw speech audio.
  arXiv  Detail & Related papers  (2023-08-11T08:03:28Z)
• MEmoBERT: Pre-training Model with Prompt-based Learning for Multimodal Emotion Recognition [118.73025093045652]
  We propose a pre-training model textbfMEmoBERT for multimodal emotion recognition. Unlike the conventional "pre-train, finetune" paradigm, we propose a prompt-based method that reformulates the downstream emotion classification task as a masked text prediction. Our proposed MEmoBERT significantly enhances emotion recognition performance.
  arXiv  Detail & Related papers  (2021-10-27T09:57:00Z)
• Multimodal Emotion Recognition with High-level Speech and Text Features [8.141157362639182]
  We propose a novel cross-representation speech model to perform emotion recognition on wav2vec 2.0 speech features. We also train a CNN-based model to recognize emotions from text features extracted with Transformer-based models. Our method is evaluated on the IEMOCAP dataset in a 4-class classification problem.
  arXiv  Detail & Related papers  (2021-09-29T07:08:40Z)
• Limited Data Emotional Voice Conversion Leveraging Text-to-Speech: Two-stage Sequence-to-Sequence Training [91.95855310211176]
  Emotional voice conversion aims to change the emotional state of an utterance while preserving the linguistic content and speaker identity. We propose a novel 2-stage training strategy for sequence-to-sequence emotional voice conversion with a limited amount of emotional speech data. The proposed framework can perform both spectrum and prosody conversion and achieves significant improvement over the state-of-the-art baselines in both objective and subjective evaluation.
  arXiv  Detail & Related papers  (2021-03-31T04:56:14Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.