Adaptive Domain-Aware Representation Learning for Speech Emotion Recognition
The DOI matches its Crossref record; 20 of the 20 references it cites were found in a registry or an index.
Abstract
Speech emotion recognition is a crucial part in human-computer interaction. However, representation learning is challenging due to much variability from speech emotion signals across diverse domains, such as gender, age, languages, and social cultural context. Many approaches focus on domain-invariant representation learning which loses the domain-specific knowledge and results in unsatisfactory speech emotion recognition across domains. In this paper, we propose an adaptive domain-aware representation learning that leverages the domain knowledge to extract domain aware features. The proposed approach applies attention model on frequency to embed the domain knowledge in the emotion representation space. Experiments demonstrate that our approach on IEMOCAP achieves the state-of-the-art performance under the same experimental conditions with WA of 73 . 02% and UA of 65 . 86% .
Overview
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
In the citing paper
“since ADARL adds a domain adaptation strategy to eliminate the discrepancy between speakers of the training data and the testing data, it obtains better performance than ATFNN on UAR.”
From Speech Emotion Recognition via an Attentive Time–Frequency Neural Network (2023) · cites this paper as [32]
In the citing paper
“it also obtains suboptimal results on UAR (64.48 % ) than ADARL (65.86 % ) [32].”
From Speech Emotion Recognition via an Attentive Time–Frequency Neural Network (2023) · cites this paper as [32]
In the citing paper
“since ADARL adds a domain adaptation strategy to eliminate the discrepancy between speakers of the training data and the testing data”
From Speech Emotion Recognition via an Attentive Time–Frequency Neural Network (2023) · cites this paper as [32]
Citations
Cites
Squeeze-and-Excitation Networks
Hu, J., Shen, L., Albanie, S., Sun, G., & Wu, E. (2020) · IEEE Transactions on Pattern Analysis and Machine Intelligence · doi:10.1109/tpami.2019.2913372 · Cited by 12,157
The Geneva Minimalistic Acoustic Parameter Set (GeMAPS) for Voice Research and Affective Computing
Eyben, F., Scherer, K. R., Schuller, B. W., Sundberg, J., Andre, E., Busso, C., Devillers, L. Y., Epps, J., Laukka, P., Narayanan, S. S., et al. (2016) · IEEE Transactions on Affective Computing · doi:10.1109/taffc.2015.2457417 · Cited by 1,818 · cited 1×
“In [3], the basic acoustic features sets, Geneva Minimalistic Acoustic Parameter Set (GeMAPS) and the extended GeMAPS are proposed.”
MethodSchuller, B., Steidl, S., Batliner, A., Vinciarelli, A., Scherer, K., Ringeval, F., Chetouani, M., Weninger, F., Eyben, F., Marchi, E., et al. (2013) · Interspeech 2013 · doi:10.21437/interspeech.2013-56 · Cited by 731 · cited 1×
“The INTERSPEECH 2013 COMPARE feature set [2] contains 6373 emotion-related features and promotes the development in SER.”
BackgroundHidden Markov model-based speech emotion recognition
Schuller, B., Rigoll, G., & Lang, M. (2003) · 2003 International Conference on Multimedia and Expo. ICME '03. Proceedings (Cat. No.03TH8698) · doi:10.1109/icme.2003.1220939 · Cited by 570 · cited 1×
“In [6], Schuller has applied Hidden Markov Model (HMM) to SER. Lee [7] has also made some progresses using decision tree.”
MethodZhang, S., Zhang, S., Huang, T., & Gao, W. (2018) · IEEE Transactions on Multimedia · doi:10.1109/tmm.2017.2766843 · Cited by 441 · cited 1×
“With the development of deep learning, LSTM [9, 4] and CNN [10, 11] , which have better ability to capture high level semantics, have become the most common classifiers.”
BackgroundEfficient emotion recognition from speech using deep learning on spectrograms
Satt, A., Rozenberg, S., & Hoory, R. (2017) · Interspeech 2017 · doi:10.21437/interspeech.2017-200 · Cited by 422 · cited 4×
“In [19], three LSTMs are used to model a common domain (domain-invariant) and two orthogonal domains (domain-specific) for text emotion classi-fication as Figure 1 (c).”
BackgroundInfluentialEmotion recognition using a hierarchical binary decision tree approach
Lee, C.-C., Mower, E., Busso, C., Lee, S., & Narayanan, S. (2011) · Speech Communication · doi:10.1016/j.specom.2011.06.004 · Cited by 355 · cited 1×
“In [6], Schuller has applied Hidden Markov Model (HMM) to SER. Lee [7] has also made some progresses using decision tree.”
MethodWöllmer, M., Eyben, F., Reiter, S., Schuller, B., Cox, C., Douglas-Cowie, E., & Cowie, R. (2008) · Interspeech 2008 · doi:10.21437/interspeech.2008-192 · Cited by 328 · cited 3×
“In [19], three LSTMs are used to model a common domain (domain-invariant) and two orthogonal domains (domain-specific) for text emotion classi-fication as Figure 1 (c).”
Continual learning of context-dependent processing in neural networks
Zeng, G., Chen, Y., Cui, B., & Yu, S. (2019) · Nature Machine Intelligence · doi:10.1038/s42256-019-0080-x · Cited by 309 · cited 2×
“Inspired by [20], we propose an adaptive domain-aware representation Learning (ADARL) in this paper, which consists of a multi-task representation learning by combining domain clas-sification and emotion recognition, and an extracting domain aware representation through the domain knowledge…”
BackgroundMethodNeumann, M., & Vu, N. T. (2017) · Interspeech 2017 · doi:10.21437/interspeech.2017-917 · Cited by 243 · cited 1×
“In addition, some studies [14, 15, 16] have benefited from multi-task learning structures to improve performance.”
Background
Cited by
SpeechFormer++: A Hierarchical Efficient Framework for Paralinguistic Speech Processing
Chen, W., Xing, X., Xu, X., Pang, J., & Du, L. (2023) · IEEE/ACM Transactions on Audio Speech and Language Processing · doi:10.1109/taslp.2023.3235194 · Cited by 117
Self-citationISNet: Individual Standardization Network for Speech Emotion Recognition
Fan, W., Xu, X., Cai, B., & Xing, X. (2022) · IEEE/ACM Transactions on Audio Speech and Language Processing · doi:10.1109/taslp.2022.3171965 · Cited by 70
Self-citationSpeech Emotion Recognition via an Attentive Time–Frequency Neural Network
Lu, C., Zheng, W., Lian, H., Zong, Y., Tang, C., Li, S., & Zhao, Y. (2023) · IEEE Transactions on Computational Social Systems · doi:10.1109/tcss.2022.3219825 · Cited by 51
“…GRU layer upon CNN layers with sequential capsules (CNN_GRU-SeqCap) [29], attention-based fully convolutional network (FCN + attention model) [23], fusion model of acoustic and linguistic features (Model-3_fusion) [30], and an adaptive domain-aware representation learning-based model (ADARL) [32].”
MethodMulti-type features separating fusion learning for Speech Emotion Recognition
Xu, X., Li, D., Zhou, Y., & Wang, Z. (2022) · Applied Soft Computing · doi:10.1016/j.asoc.2022.109648 · Cited by 35
Self-citationMGAT: Multi-Granularity Attention Based Transformers for Multi-Modal Emotion Recognition
Fan, W., Xing, X., Cai, B., & Xu, X. (2023) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp49357.2023.10095855 · Cited by 13
“It addresses emotional asynchrony and modality misalignment through a multi-granularity attention (MGA) mechanism during the interaction process.”
BackgroundSelf-citationImproving Speaker-Independent Speech Emotion Recognition using Dynamic Joint Distribution Adaptation
Lu, C., Zong, Y., Lian, H., Zhao, Y. D., Schuller, B. W., & Zheng, W. (2024) · IEEE International Conference on Acoustics Speech and Signal Processing · doi:10.1109/icassp48485.2024.10447452 · Cited by 15
“Fortunately, recent SER studies have demonstrated that Domain Adaptation (DA) holds promise in mitigating the domain/speaker bias in SER [7], [8], [9].”
BackgroundMethodInfluentialFan, W., Xu, X., Guohua, Z., Deng, X., & Xing, X. (2025) · Speech Communication · doi:10.1016/j.specom.2025.103198 · Cited by 13
Self-citationDaneshfar, F., & Kabudian, S. J. (2021) · arXiv (Cornell University) · doi:10.48550/arxiv.2111.07094 · Cited by 9
“And finally in the model proposed by [29] an adaptive domain-aware representation learning method is used to better identify and extract domain-dependent features.”
MethodFan, W., Xu, X., Liu, F., & Xing, X. (2025) · Expert Systems with Applications · doi:10.1016/j.eswa.2025.127669 · Cited by 5
Self-citationGeneralizable Dynamic Domain Adaptation for Speaker-Independent Speech Emotion Recognition
Lu, C., Zong, Y., Lian, H., Li, S., Zhao, Y. D., Schuller, B. W., & Zheng, W. (2025) · IEEE Transactions on Audio Speech and Language Processing · doi:10.1109/taslpro.2025.3636055 · Cited by 4
“In the experiments on IEMOCAP, we select some popular comparison methods,namelyDNN-HMM_SGMM-Ali[35],aCNN+LSTM Model[36],CNN_GRU-SeqCap[37],aFCN + Attention Model [38], Model-3 Fusion [39], ADARL [40], LGFA [41], and ATFNN [42].”
Xi, Y., Song, Y., Dai, L.-R., & Liu, L. (2023) · Communications in computer and information science · doi:10.1007/978-981-99-2401-1_16
Research on Speech Emotion Recognition Base on SVM
Zhang, L., Ablimit, M., & Hamdulla, A. (2022) · International Conference on Intelligent Transportation, Big Data & Smart City (ICITBS) · doi:10.1109/icitbs55627.2022.00060
DA-KWFormer: A Domain Adaptation Network with K-Weight Transformer for Speech Emotion Recognition
Xu, F., Yang, H., & Hu, Y. (2024) · Communications in computer and information science · doi:10.1007/978-981-96-1045-7_29 · Cited by 1
Emotion transfer in audio using mel-cepstral representation and CycleGANs
Han, G., Zhao, J., & Zhou, Y. (2025) · Journal of King Saud University - Computer and Information Sciences · doi:10.1007/s44443-025-00082-7 · Cited by 1
“In addition, compared with the content encoder, the emotion encoder also adds a Temporal Avg Pool layer (Fan et al. 2020).”
Metadata
- Authors
- Fan, W., Xu, X., Xing, X., & Huang, D.
- Venue
- Interspeech 2020
- Publisher
- ISCA
- Published
- 2020-10-25
- Pages
- 4089-4093
- ISSN
- 2308-457X (ISSN-L)
- OpenAlex
- W3096039514
- Cited by
- 14
- References
- 20
- Citations by year
- 2025: 4 · 2024: 2 · 2023: 3 · 2022: 4 · 2021: 1
- FWCI
- 1.95
- Type
- conference paper
- Language
- en
- Open access
- closed
- Locations
- Interspeech (USB) (publishedVersion)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Emotion and Mood Recognition
- Physical Sciences › Computer Science › Artificial Intelligence › Speech Recognition and Synthesis
- Physical Sciences › Computer Science › Signal Processing › Speech and Audio Processing
- Keywords
- speech emotion recognition, domain adaptation, representation learning
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
Data sources
Bibliographic data: Semantic Scholar, OpenAlex, Crossref.
Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).
Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.