Abinay Reddy Naini
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Abinay Reddy Naini has a total of 11 co-authors. Here are the top co-authors by number of shared publications:
- Carlos Busso
Language Technologies Institute, Carnegie Mellon University, Pittsburgh, Pennsylvania, USA
2 shared publications - Ali N. Salman
The University of Texas at Dallas, Richardson, TX, USA
1 shared publication - Huang-Cheng Chou
Department of Electrical Engineering, National Tsing Hua University, Hsinchu City, Taiwan
1 shared publication - Kusha Sridhar
Accenture LLP, Mountain View, CA, USA
1 shared publication - Lucas Goncalves
The University of Texas at Dallas, Richardson, TX, USA
1 shared publication - Luz Martinez-Lucas
Erik Jonsson School of Engineering and Computer Science, University of Texas at Dallas, Richardson, TX, USA
1 shared publication
Latest Publications
- Open Access
- 6 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Affective Computing
- Vol. PP (3) , 1-14
Ordinal-Based Speech Emotion Recognition Leveraging Consecutive Trends: A Curriculum-Driven Preference Learning Approach
In the domain of human-computer interaction, accurately ranking speech samples by their expressive content is crucial for enhancing emotional understanding and response mechanisms. Traditional methods, relying on absolute attribute scores annotated by multiple raters, face challenges in label reliability for training preference learning frameworks. This paper introduces a novel preference learning approach that leverages the inherent trends in raters' annotations of consecutive samples, circumventing the inconsistencies associated with absolute scoring. By focusing on annotation trends within evaluation sessions, our method significantly reduces the ambiguity in preference labels, providing a more robust foundation for preference learning. Our approach uses curriculum learning (CL) to further refine this approach. Common methods to create relative labels from absolute scores require setting fixed margins to establish preferences. We can avoid fixed thresholds by using CL, where strict margins are used early in training and progressively relaxed. CL's structured progression inherently aligns with the sequential nature of our preference learning model, enhancing the model's ability to discern subtle emotional cues within speech samples. Comparative analysis across diverse corpora, including the MSP-Podcast, IEMOCAP, and BIIC-Podcast corpora, reveals that this combination of trend-based preference learning and curriculum learning not only yields superior performance within individual datasets but also demonstrates significant cross-corpus generalizability. This cross-corpus efficacy underscores the robustness and adaptability of our methodology, marking a substantial advancement in the field of emotion recognition and human-computer interaction.
- 6 April 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Affective Computing
- Vol. PP (3) , 1-14
Open AccessOrdinal-Based Speech Emotion Recognition Leveraging Consecutive Trends: A Curriculum-Driven Preference Learning Approach
In the domain of human-computer interaction, accurately ranking speech samples by their expressive content is crucial for enhancing emotional understanding and response mechanisms. Traditional methods, relying on absolute attribute scores annotated by multiple raters, face challenges in label reliability for training preference learning frameworks. This paper introduces a novel preference learning approach that leverages the inherent trends in raters' annotations of consecutive samples, circumventing the inconsistencies associated with absolute scoring. By focusing on annotation trends within evaluation sessions, our method significantly reduces the ambiguity in preference labels, providing a more robust foundation for preference learning. Our approach uses curriculum learning (CL) to further refine this approach. Common methods to create relative labels from absolute scores require setting fixed margins to establish preferences. We can avoid fixed thresholds by using CL, where strict margins are used early in training and progressively relaxed. CL's structured progression inherently aligns with the sequential nature of our preference learning model, enhancing the model's ability to discern subtle emotional cues within speech samples. Comparative analysis across diverse corpora, including the MSP-Podcast, IEMOCAP, and BIIC-Podcast corpora, reveals that this combination of trend-based preference learning and curriculum learning not only yields superior performance within individual datasets but also demonstrates significant cross-corpus generalizability. This cross-corpus efficacy underscores the robustness and adaptability of our methodology, marking a substantial advancement in the field of emotion recognition and human-computer interaction.
- Open Access
- 30 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Affective Computing
- Vol. PP (3) , 1-19
The MSP-Podcast Corpus
The availability of large, high-quality emotional speech databases is essential for advancing speech emotion recognition (SER) in real-world scenarios. However, many existing databases face limitations in size, emotional balance, and speaker diversity. This study describes the MSP-Podcast corpus, summarizing our ten-year effort. The corpus consists of over 400 hours of diverse audio samples from various audio sharing websites, all of which have Common Licenses that permit the distribution of the corpus. We annotate the corpus with rich emotional labels, including primary (single dominant emotion) and secondary (multiple emotions perceived in the audio) emotional categories, as well as emotional attributes for valence, arousal, and dominance. At least five raters annotate these emotional labels. The corpus also has speaker identification for most samples, and human transcriptions of the lexical content of the sentences for the entire corpus. The data collection protocol includes a machine learning-driven pipeline for selecting emotionally diverse recordings, ensuring a balanced and varied representation of emotions across speakers and environments. The resulting database provides a comprehensive, high-quality resource, better suited for advancing SER systems in practical, real-world scenarios.
- 30 March 2026
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Transactions on Affective Computing
- Vol. PP (3) , 1-19
Open AccessThe MSP-Podcast Corpus
The availability of large, high-quality emotional speech databases is essential for advancing speech emotion recognition (SER) in real-world scenarios. However, many existing databases face limitations in size, emotional balance, and speaker diversity. This study describes the MSP-Podcast corpus, summarizing our ten-year effort. The corpus consists of over 400 hours of diverse audio samples from various audio sharing websites, all of which have Common Licenses that permit the distribution of the corpus. We annotate the corpus with rich emotional labels, including primary (single dominant emotion) and secondary (multiple emotions perceived in the audio) emotional categories, as well as emotional attributes for valence, arousal, and dominance. At least five raters annotate these emotional labels. The corpus also has speaker identification for most samples, and human transcriptions of the lexical content of the sentences for the entire corpus. The data collection protocol includes a machine learning-driven pipeline for selecting emotionally diverse recordings, ensuring a balanced and varied representation of emotions across speakers and environments. The resulting database provides a comprehensive, high-quality resource, better suited for advancing SER systems in practical, real-world scenarios.