Speech emotion recognition (SER) systems often exhibit gender bias. However, the effectiveness and robustness of existing debiasing methods in such multi-label scenarios remain underexplored. To address this gap, we present EMO-Debias—a large-scale comparison of 13 debiasing methods applied to multi-label SER. Our study encompasses techniques from pre-processing, regularization, adversarial learning, biased learners, and distributionally robust optimization. Experiments conducted on acted and naturalistic emotion datasets, using WavLM and XLSR representations, evaluate each method under conditions of gender imbalance. Our analysis quantifies the trade-offs between fairness and accuracy, identifying which approaches consistently reduce gender performance gaps without compromising overall model performance. The findings provide actionable insights for selecting effective debiasing strategies and highlight the impact of dataset distributions.

Speech emotions play a crucial role in human-computer interaction, shaping engagement and context-aware communication. Despite recent advances in spoken dialogue systems, a holistic system for evaluating emotional reasoning is still lacking. To address this, we introduce EMO-Reasoning, a benchmark for assessing emotional coherence in dialogue systems. It leverages a curated dataset generated via text-to-speech to simulate diverse emotional states, overcoming the scarcity of emotional speech data. We further propose the Cross-turn Emotion Reasoning Score to assess the emotion transitions in multi-turn dialogues. Evaluating seven dialogue systems through continuous, categorical, and perceptual metrics, we show that our framework effectively detects emotional inconsistencies, providing insights for improving current dialogue systems. By releasing a systematic evaluation benchmark, we aim to advance emotion-aware spoken dialogue modeling toward more natural and adaptive interactions.

Bias in speech emotion recognition (SER) systems often stems from spurious correlations between speaker characteristics and emotional labels, leading to unfair predictions across demographic groups. Many existing debiasing methods require model-specific changes or demographic annotations, limiting their practical use. We present COVADA, a Confidence-Oriented Voice Augmentation Debiasing Approach that mitigates bias without modifying model architecture or relying on demographic information. CO-VADA identifies training samples that reflect bias patterns present in the training data and then applies voice conversion to alter irrelevant attributes and generate samples. These augmented samples introduce speaker variations that differ from dominant patterns in the data, guiding the model to focus more on emotion-relevant features. Our framework is compatible with various SER models and voice conversion tools, making it a scalable and practical solution for improving fairness in SER systems.

Speech Emotion Recognition (SER) systems rely on speech input and emotional labels annotated by humans. However, various emotion databases collect perceptional evaluations in different ways. For instance, the IEMOCAP dataset uses video clips with sounds for annotators to provide their emotional perceptions. However, the most significant English emotion dataset, the MSP-PODCAST, only provides speech for raters to choose the emotional ratings. Nevertheless, using speech as input is the standard approach to training SER systems. Therefore, the open question is the emotional labels elicited by which scenarios are the most effective for training SER systems. We comprehensively compare the effectiveness of SER systems trained with labels elicited by different modality stimuli and evaluate the SER systems on various testing conditions. Also, we introduce an all-inclusive label that combines all labels elicited by various modalities. We show that using labels elicited by voice-only stimuli for training yields better performance on the test set, whereas labels elicited by voice-only stimuli.

Speech Emotion Recognition (SER) is a crucial component in developing general-purpose AI agents capable of natural human-computer interaction. However, building robust multilingual SER systems remains challenging due to the scarcity of labeled data in languages other than English and Chinese. In this paper, we propose an approach to enhance SER performance in low SER resource languages by leveraging data from high-resource languages. Specifically, we employ expressive Speech-to-Speech translation (S2ST) combined with a novel bootstrapping data selection pipeline to generate labeled data in the target language. Extensive experiments demonstrate that our method is both effective and generalizable across different upstream models and languages. Our results suggest that this approach can facilitate the development of more scalable and robust multilingual SER systems. Our code is available at: https://github.com/hsi-che-lin/Improve-SER-via-S2ST

When selecting test data for subjective tasks, most studies define ground truth labels using aggregation methods such as the majority or plurality rules. These methods discard data points without consensus, making the test set easier than practical tasks where a prediction is needed for each sample. However, the discarded data points often express ambiguous cues that elicit coexisting traits perceived by annotators. This paper addresses the importance of considering all the annotations and samples in the data, highlighting that only showing the model's performance on an incomplete test set selected by using the majority or plurality rules can lead to bias in the models’ performances. We focus onspeech-emotion recognition(SER) tasks. We observe that traditional aggregation rules have a data loss ratio ranging from 5.63% to 89.17%. From this observation, we propose a flexible method named the all-inclusive aggregation rule to evaluate SER systems on the complete test data. We contrast traditional single-label formulations with a multi-label formulation to consider the coexistence of emotions. We show that training an SER model with the data selected by the all-inclusive aggregation rule shows consistently higher macro-F1 scores when tested in the entire test set, including ambiguous samples without agreement.

Speech Emotion Recognition (SER) faces a distinct challenge compared to other speech-related tasks because the annotations will show the subjective emotional perceptions of different annotators. Previous SER studies often view the subjectivity of emotion perception as noise by using the majority rule or plurality rule to obtain the consensus labels. However, these standard approaches overlook the valuable information of labels that do not agree with the consensus and make it easier for the test set. Emotion perception can have co-occurring emotions in realistic conditions, and it is unnecessary to regard the disagreement between raters as noise. To bridge the SER into a multi-label task, we introduced an “all-inclusive rule,” which considers all available data, ratings, and distributional labels as multi-label targets and a complete test set. We demonstrated that models trained with multi-label targets generated by the proposed AR outperform conventional single-label methods across incomplete and complete test sets.

Speech emotion recognition (SER) is an essential technology for human-computer interaction systems. However, the previous study reveals that 80.77% of SER papers yield results that cannot be reproduced on the well-known IEMOCAP dataset. The main reason for reproducibility challenges is that the database did not provide standard data splits (e.g., train, development, and test sets). Prior papers could define its partition, but they did not provide details of the partition or source code for processing the partition. Therefore, this work aims to make SER open and reproducible to everyone. We develop the EMO-SUPERB, shorted for EMOtion Speech Universal PERformance Benchmark, including a user-friendly codebase to leverage 16 state-of-the-art (SOTA) speech self-supervised learning models for exhaustive evaluation plus one SOTA SER model across 6 open-source SER datasets in English and Chinese. We make all resources open-source to facilitate future developments in SER. Researchers can easily upload their systems or datasets to EMO-SUPERB, and we name the project “Open-Emotion”.

The decision of ground truth for speech emotion recognition (SER) is still a critical issue in affective computing tasks. Previous studies on emotion recognition often rely on consensus labels after aggregating the classes selected by multiple annotators. It is common for a perceptual evaluation conducted to annotate emotional corpora to include the class “other,” allowing the annotators the opportunity to describe the emotion with their own words. This practice provides valuable emotional information, which, however, is ignored in most emotion recognition studies. This paper utilizes easy-accessed natural language processing toolkits to mine the sentiment of these typed descriptions, enriching and maximizing the information obtained from the annotators. The polarity information is combined with primary and secondary annotations provided by individual evaluators under a label distribution framework, creating a complete representation of the emotional content of the spoken sentences. Finally, we train multitask learning SER models with existing learning methods (soft-label, multi-label, and distribution-label) to show the performance of the novel ground truth in the MSP-Podcast corpus.

Emotion perception is subjective and vary with respect to each individual due to the natural bias of human, such as gender, culture, and age. Conventionally, emotion recognition relies on the consensus, e.g., majority of annotations (hard label) or the distribution of annotations (soft label), and do not include rater-specific model. In this paper, we propose a joint learning methodology that simultaneously considers the label uncertainty and annotator idiosyncrasy using hard and soft emotion label annotation accompanying with individual and crowd annotator modeling. Our proposed model achieves unweighted average recall (UAR) 61.48% on the benchmark emotion corpus. Further analyses reveal that emotion perception is indeed rater-dependent, using the hard label and soft emotion distribution provides complementary affect modeling information, and finally joint learning of subjective emotion perception and individual rater model provides the best discriminative power.

The increasing availability of large-scale emotion corpus along with the advancement in emotion recognition algorithms have enabled the emergence of next-generation human-machine interfaces. This paper describes a newly-collected multimodal corpus, i.e., the NTHU-NTUA Chinese Interactive Emotion Corpus (NNIME). The database is a result of the collaborative work between engineers and drama experts. This database includes recordings of 44 subjects engaged in spontaneous dyadic spoken interactions. The multimodal data includes approximately 11-hour worth of audio, video, and electrocardiogram data recorded continuously and synchronously. The database is also completed with a rich set of emotion annotations on discrete and continuous-in-time annotation by a total of 49 annotators. Thees emotion annotations include a diverse perspectives: peer-report, director-report, self-report, and observer-report. This carefully-engineered data collection and annotation processes provide an additional valuable resource to quantify and investigate various aspects of affective phenomenon and human communication. To our best knowledge, the NNIME is one of the few large-scale Chinese affective dyadic interaction database that have been systematically collected, organized, and to be publicly-released to the research community.