Emotion Recognition in Conversation (ERC) is a practical and challenging task. This paper proposes a novel multimodal approach, the Long-Short Distance Graph Neural Network (LSDGNN). Based on the Directed Acyclic Graph (DAG), it constructs a long-distance graph neural network and a short-distance graph neural network to obtain multimodal features of distant and nearby utterances, respectively. To ensure that long- and short-distance features are as distinct as possible in representation while enabling mutual influence between the two modules, we employ a Differential Regularizer and incorporate a BiAffine Module to facilitate feature interaction. In addition, we propose an Improved Curriculum Learning (ICL) to address the challenge of data imbalance. By computing the similarity between different emotions to emphasize the shifts in similar emotions, we design a "weighted emotional shift" metric and develop a difficulty measurer, enabling a training process that prioritizes learning easy samples before harder ones. Experimental results on the IEMOCAP and MELD datasets demonstrate that our model outperforms existing benchmarks.
Recent research has focused on applying speech large language model (SLLM) to improve speech emotion recognition (SER). However, the inherently high frame rate in speech modality severely limits the signal processing and understanding capabilities of SLLM. For example, a SLLM with a 4K context window can only process 80 seconds of audio at 50Hz feature sampling rate before reaching its capacity limit. Input token compression methods used in SLLM overlook the continuity and inertia of emotions across multiple conversation turns. This paper proposes a Dynamic Parameter Memory (DPM) mechanism with contextual semantics and sentence-level emotion encoding, enabling processing of unlimited-length audio with limited context windows in SLLM. Specifically, DPM progressively encodes sentence-level information and emotions into a temporary LoRA module during inference to effectively "memorize" the contextual information. We trained an emotion SLLM as a backbone and incorporated our DPM into inference for emotion recognition in conversation (ERC). Experimental results on the IEMOCAP dataset show that DPM significantly improves the emotion recognition capabilities of SLLM when processing long audio sequences, achieving state-of-the-art performance.
Remote photoplethysmography (rPPG) is a non-contact technique for measuring human physiological signals. Due to its convenience and non-invasiveness, it has demonstrated broad application potential in areas such as health monitoring and emotion recognition. In recent years, the release of numerous public datasets has significantly advanced the performance of rPPG algorithms under ideal lighting conditions. However, the effectiveness of current rPPG methods in realistic nighttime scenarios with dynamic lighting variations remains largely unknown. Moreover, there is a severe lack of datasets specifically designed for such challenging environments, which has substantially hindered progress in this area of research. To address this gap, we present and release a large-scale rPPG dataset collected under dynamic lighting conditions at night, named DLCN. The dataset comprises approximately 13 hours of video data and corresponding synchronized physiological signals from 98 participants, covering four representative nighttime lighting scenarios. DLCN offers high diversity and realism, making it a valuable resource for evaluating algorithm robustness in complex conditions. Built upon the proposed Happy-rPPG Toolkit, we conduct extensive experiments and provide a comprehensive analysis of the challenges faced by state-of-the-art rPPG methods when applied to DLCN. The dataset and code are publicly available at https://github.com/dalaoplan/Happp-rPPG-Toolkit.
Infant cry emotion recognition is crucial for parenting and medical applications. It faces many challenges, such as subtle emotional variations, noise interference, and limited data. The existing methods lack the ability to effectively integrate multi-scale features and temporal-frequency relationships. In this study, we propose a method for infant cry emotion recognition using an improved Emphasized Channel Attention, Propagation and Aggregation in Time Delay Neural Network (ECAPA-TDNN) with both multi-scale feature fusion and attention enhancement. Experiments on a public dataset show that the proposed method achieves accuracy of 82.20%, number of parameters of 1.43 MB and FLOPs of 0.32 Giga. Moreover, our method has advantage over the baseline methods in terms of accuracy. The code is at https://github.com/kkpretend/IETMA.
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.
Recent advancements in Multimodal Emotion Recognition (MER) face challenges in addressing both modality missing and Out-Of-Distribution (OOD) data simultaneously. Existing methods often rely on specific models or introduce excessive parameters, which limits their practicality. To address these issues, we propose a novel robust MER framework, Causal Inference Distiller (CIDer), and introduce a new task, Random Modality Feature Missing (RMFM), to generalize the definition of modality missing. CIDer integrates two key components: a Model-Specific Self-Distillation (MSSD) module and a Model-Agnostic Causal Inference (MACI) module. MSSD enhances robustness under the RMFM task through a weight-sharing self-distillation approach applied across low-level features, attention maps, and high-level representations. Additionally, a Word-level Self-aligned Attention Module (WSAM) reduces computational complexity, while a Multimodal Composite Transformer (MCT) facilitates efficient multimodal fusion. To tackle OOD challenges, MACI employs a tailored causal graph to mitigate label and language biases using a Multimodal Causal Module (MCM) and fine-grained counterfactual texts. Notably, MACI can independently enhance OOD generalization with minimal additional parameters. Furthermore, we also introduce the new repartitioned MER OOD datasets. Experimental results demonstrate that CIDer achieves robust performance in both RMFM and OOD scenarios, with fewer parameters and faster training compared to state-of-the-art methods. The implementation of this work is publicly accessible at https://github.com/gw-zhong/CIDer.
Speech Emotion Recognition (SER) is crucial for improving human-computer interaction. Despite strides in monolingual SER, extending them to build a multilingual system remains challenging. Our goal is to train a single model capable of multilingual SER by distilling knowledge from multiple teacher models. To address this, we introduce a novel language-aware multi-teacher knowledge distillation method to advance SER in English, Finnish, and French. It leverages Wav2Vec2.0 as the foundation of monolingual teacher models and then distills their knowledge into a single multilingual student model. The student model demonstrates state-of-the-art performance, with a weighted recall of 72.9 on the English dataset and an unweighted recall of 63.4 on the Finnish dataset, surpassing fine-tuning and knowledge distillation baselines. Our method excels in improving recall for sad and neutral emotions, although it still faces challenges in recognizing anger and happiness.
Discourse parsing is an important task useful for NLU applications such as summarization, machine comprehension, and emotion recognition. The current discourse parsing datasets based on conversations consists of written English dialogues restricted to a single domain. In this resource paper, we introduce CoMuMDR: Code-mixed Multi-modal Multi-domain corpus for Discourse paRsing in conversations. The corpus (code-mixed in Hindi and English) has both audio and transcribed text and is annotated with nine discourse relations. We experiment with various SoTA baseline models; the poor performance of SoTA models highlights the challenges of multi-domain code-mixed corpus, pointing towards the need for developing better models for such realistic settings.
Although speech emotion recognition (SER) has advanced significantly with deep learning, annotation remains a major hurdle. Human annotation is not only costly but also subject to inconsistencies annotators often have different preferences and may lack the necessary contextual knowledge, which can lead to varied and inaccurate labels. Meanwhile, Large Language Models (LLMs) have emerged as a scalable alternative for annotating text data. However, the potential of LLMs to perform emotional speech data annotation without human supervision has yet to be thoroughly investigated. To address these problems, we apply GPT-4o to annotate a multimodal dataset collected from the sitcom Friends, using only textual cues as inputs. By crafting structured text prompts, our methodology capitalizes on the knowledge GPT-4o has accumulated during its training, showcasing that it can generate accurate and contextually relevant annotations without direct access to multimodal inputs. Therefore, we propose MELT, a multimodal emotion dataset fully annotated by GPT-4o. We demonstrate the effectiveness of MELT by fine-tuning four self-supervised learning (SSL) backbones and assessing speech emotion recognition performance across emotion datasets. Additionally, our subjective experiments\' results demonstrate a consistence performance improvement on SER.
Opinion mining plays a vital role in analysing user feedback and extracting insights from textual data. While most research focuses on sentiment polarity (e.g., positive, negative, neutral), fine-grained emotion classification in app reviews remains underexplored. This paper addresses this gap by identifying and addressing the challenges and limitations in fine-grained emotion analysis in the context of app reviews. Our study adapts Plutchik's emotion taxonomy to app reviews by developing a structured annotation framework and dataset. Through an iterative human annotation process, we define clear annotation guidelines and document key challenges in emotion classification. Additionally, we evaluate the feasibility of automating emotion annotation using large language models, assessing their cost-effectiveness and agreement with human-labelled data. Our findings reveal that while large language models significantly reduce manual effort and maintain substantial agreement with human annotators, full automation remains challenging due to the complexity of emotional interpretation. This work contributes to opinion mining by providing structured guidelines, an annotated dataset, and insights for developing automated pipelines to capture the complexity of emotions in app reviews.
The ability to handle various emotion labels without dedicated training is crucial for building adaptable Emotion Recognition (ER) systems. Conventional ER models rely on training using fixed label sets and struggle to generalize beyond them. On the other hand, Large Language Models (LLMs) have shown strong zero-shot ER performance across diverse label spaces, but their scale limits their use on edge devices. In this work, we propose a contrastive distillation framework that transfers rich emotional knowledge from LLMs into a compact model without the use of human annotations. We use GPT-4 to generate descriptive emotion annotations, offering rich supervision beyond fixed label sets. By aligning text samples with emotion descriptors in a shared embedding space, our method enables zero-shot prediction on different emotion classes, granularity, and label schema. The distilled model is effective across multiple datasets and label spaces, outperforming strong baselines of similar size and approaching GPT-4's zero-shot performance, while being over 10,000 times smaller.
Electroencephalography-based Emotion Recognition (EEG-ER) has become a growing research area in recent years. Analyzing 216 papers published between 2018 and 2023, we uncover that the field lacks a unified evaluation protocol, which is essential to fairly define the state of the art, compare new approaches and to track the field's progress. We report the main inconsistencies between the used evaluation protocols, which are related to ground truth definition, evaluation metric selection, data splitting types (e.g., subject-dependent or subject-independent) and the use of different datasets. Capitalizing on this state-of-the-art research, we propose a unified evaluation protocol, EEGain (https://github.com/EmotionLab/EEGain), which enables an easy and efficient evaluation of new methods and datasets. EEGain is a novel open source software framework, offering the capability to compare - and thus define - state-of-the-art results. EEGain includes standardized methods for data pre-processing, data splitting, evaluation metrics, and the ability to load the six most relevant datasets (i.e., AMIGOS, DEAP, DREAMER, MAHNOB-HCI, SEED, SEED-IV) in EEG-ER with only a single line of code. In addition, we have assessed and validated EEGain using these six datasets on the four most common publicly available methods (EEGNet, DeepConvNet, ShallowConvNet, TSception). This is a significant step to make research on EEG-ER more reproducible and comparable, thereby accelerating the overall progress of the field.
Speech emotion recognition (SER) in naturalistic settings remains a challenge due to the intrinsic variability, diverse recording conditions, and class imbalance. As participants in the Interspeech Naturalistic SER Challenge which focused on these complexities, we present Abhinaya, a system integrating speech-based, text-based, and speech-text models. Our approach fine-tunes self-supervised and speech large language models (SLLM) for speech representations, leverages large language models (LLM) for textual context, and employs speech-text modeling with an SLLM to capture nuanced emotional cues. To combat class imbalance, we apply tailored loss functions and generate categorical decisions through majority voting. Despite one model not being fully trained, the Abhinaya system ranked 4th among 166 submissions. Upon completion of training, it achieved state-of-the-art performance among published results, demonstrating the effectiveness of our approach for SER in real-world conditions.
In our prior works, we introduced a scalable streaming speech synthesis model, CosyVoice 2, which integrates a large language model (LLM) and a chunk-aware flow matching (FM) model, and achieves low-latency bi-streaming speech synthesis and human-parity quality. Despite these advancements, CosyVoice 2 exhibits limitations in language coverage, domain diversity, data volume, text formats, and post-training techniques. In this paper, we present CosyVoice 3, an improved model designed for zero-shot multilingual speech synthesis in the wild, surpassing its predecessor in content consistency, speaker similarity, and prosody naturalness. Key features of CosyVoice 3 include: 1) A novel speech tokenizer to improve prosody naturalness, developed via supervised multi-task training, including automatic speech recognition, speech emotion recognition, language identification, audio event detection, and speaker analysis. 2) A new differentiable reward model for post-training applicable not only to CosyVoice 3 but also to other LLM-based speech synthesis models. 3) Dataset Size Scaling: Training data is expanded from ten thousand hours to one million hours, encompassing 9 languages and 18 Chinese dialects across various domains and text formats. 4) Model Size Scaling: Model parameters are increased from 0.5 billion to 1.5 billion, resulting in enhanced performance on our multilingual benchmark due to the larger model capacity. These advancements contribute significantly to the progress of speech synthesis in the wild. We encourage readers to listen to the demo at https://funaudiollm.github.io/cosyvoice3.
Emotion plays a fundamental role in human interaction, and therefore systems capable of identifying emotions in speech are crucial in the context of human-computer interaction. Speech emotion recognition (SER) is a challenging problem, particularly in natural speech and when the available data is imbalanced across emotions. This paper presents our proposed system in the context of the 2025 Speech Emotion Recognition in Naturalistic Conditions Challenge. Our proposed architecture leverages cross-modality, utilizing cross-modal attention to fuse representations from different modalities. To address class imbalance, we employed two training designs: (i) weighted crossentropy loss (WCE); and (ii) WCE with an additional neutralexpressive soft margin loss and balancing. We trained a total of 12 multimodal models, which were ensembled using a balanced stacking model. Our proposed system achieves a MacroF1 score of 0.4094 and an accuracy of 0.4128 on 8-class speech emotion recognition.
With the rapid advancement of global digitalization, users from different countries increasingly rely on social media for information exchange. In this context, multilingual multi-label emotion detection has emerged as a critical research area. This study addresses SemEval-2025 Task 11: Bridging the Gap in Text-Based Emotion Detection. Our paper focuses on two sub-tracks of this task: (1) Track A: Multi-label emotion detection, and (2) Track B: Emotion intensity. To tackle multilingual challenges, we leverage pre-trained multilingual models and focus on two architectures: (1) a fine-tuned BERT-based classification model and (2) an instruction-tuned generative LLM. Additionally, we propose two methods for handling multi-label classification: the base method, which maps an input directly to all its corresponding emotion labels, and the pairwise method, which models the relationship between the input text and each emotion category individually. Experimental results demonstrate the strong generalization ability of our approach in multilingual emotion recognition. In Track A, our method achieved Top 4 performance across 10 languages, ranking 1st in Hindi. In Track B, our approach also secured Top 5 performance in 7 languages, highlighting its simplicity and effectiveness\footnote{Our code is available at https://github.com/yingjie7/mlingual_multilabel_emo_detection.
Facial Emotion Analysis (FEA) plays a crucial role in visual affective computing, aiming to infer a person's emotional state based on facial data. Scientifically, facial expressions (FEs) result from the coordinated movement of facial muscles, which can be decomposed into specific action units (AUs) that provide detailed emotional insights. However, traditional methods often struggle with limited interpretability, constrained generalization and reasoning abilities. Recently, Multimodal Large Language Models (MLLMs) have shown exceptional performance in various visual tasks, while they still face significant challenges in FEA due to the lack of specialized datasets and their inability to capture the intricate relationships between FEs and AUs. To address these issues, we introduce a novel FEA Instruction Dataset that provides accurate and aligned FE and AU descriptions and establishes causal reasoning relationships between them, followed by constructing a new benchmark, FEABench. Moreover, we propose FEALLM, a novel MLLM architecture designed to capture more detailed facial information, enhancing its capability in FEA tasks. Our model demonstrates strong performance on FEABench and impressive generalization capability through zero-shot evaluation on various datasets, including RAF-DB, AffectNet, BP4D, and DISFA, showcasing its robustness and effectiveness in FEA tasks. The dataset and code will be available at https://github.com/953206211/FEALLM.
Emotion recognition is essential across numerous fields, including medical applications and brain-computer interface (BCI). Emotional responses include behavioral reactions, such as tone of voice and body movement, and changes in physiological signals, such as the electroencephalogram (EEG). The latter are involuntary, thus they provide a reliable input for identifying emotions, in contrast to the former which individuals can consciously control. These signals reveal true emotional states without intentional alteration, thus increasing the accuracy of emotion recognition models. However, multimodal deep learning methods from physiological signals have not been significantly investigated. In this paper, we introduce PHemoNet, a fully hypercomplex network for multimodal emotion recognition from physiological signals. In detail, the architecture comprises modality-specific encoders and a fusion module. Both encoders and fusion modules are defined in the hypercomplex domain through parameterized hypercomplex multiplications (PHMs) that can capture latent relations between the different dimensions of each modality and between the modalities themselves. The proposed method outperforms current state-of-the-art models on the MAHNOB-HCI dataset in classifying valence and arousal using electroencephalograms (EEGs) and peripheral physiological signals. The code for this work is available at https://github.com/ispamm/MHyEEG.
Emotion understanding in videos aims to accurately recognize and interpret individuals' emotional states by integrating contextual, visual, textual, and auditory cues. While Large Multimodal Models (LMMs) have demonstrated significant progress in general vision-language (VL) tasks, their performance in emotion-specific scenarios remains limited. Moreover, fine-tuning LMMs on emotion-related tasks often leads to catastrophic forgetting, hindering their ability to generalize across diverse tasks. To address these challenges, we present Emotion-Qwen, a tailored multimodal framework designed to enhance both emotion understanding and general VL reasoning. Emotion-Qwen incorporates a sophisticated Hybrid Compressor based on the Mixture of Experts (MoE) paradigm, which dynamically routes inputs to balance emotion-specific and general-purpose processing. The model is pre-trained in a three-stage pipeline on large-scale general and emotional image datasets to support robust multimodal representations. Furthermore, we construct the Video Emotion Reasoning (VER) dataset, comprising more than 40K bilingual video clips with fine-grained descriptive annotations, to further enrich Emotion-Qwen's emotional reasoning capability. Experimental results demonstrate that Emotion-Qwen achieves state-of-the-art performance on multiple emotion recognition benchmarks, while maintaining competitive results on general VL tasks. Code and models are available at https://github.com/24DavidHuang/Emotion-Qwen.
The fusion technique is the key to the multimodal emotion recognition task. Recently, cross-modal attention-based fusion methods have demonstrated high performance and strong robustness. However, cross-modal attention suffers from redundant features and does not capture complementary features well. We find that it is not necessary to use the entire information of one modality to reinforce the other during cross-modal interaction, and the features that can reinforce a modality may contain only a part of it. To this end, we design an innovative Transformer-based Adaptive Cross-modal Fusion Network (TACFN). Specifically, for the redundant features, we make one modality perform intra-modal feature selection through a self-attention mechanism, so that the selected features can adaptively and efficiently interact with another modality. To better capture the complementary information between the modalities, we obtain the fused weight vector by splicing and use the weight vector to achieve feature reinforcement of the modalities. We apply TCAFN to the RAVDESS and IEMOCAP datasets. For fair comparison, we use the same unimodal representations to validate the effectiveness of the proposed fusion method. The experimental results show that TACFN brings a significant performance improvement compared to other methods and reaches the state-of-the-art. All code and models could be accessed from https://github.com/shuzihuaiyu/TACFN.