Sat, Oct 17, 2026
| فارسی
1- PhD Student of Computer Engineering, Department of Computer Engineering, SR.C., Islamic Azad University, Tehran, Iran
2- Professor of Electrical and Computer Engineering, Department of Medical Radiation, Amirkabir University of Technology, Tehran, Iran
3- Associate Professor of Biomedical Engineering, Department of Medical Sciences, Tehran University, Tehran, Iran
4- Assistant Professor of Computer Engineering, Department of Computer Engineering, SR.C., Islamic Azad University, Tehran, Iran
2- Professor of Electrical and Computer Engineering, Department of Medical Radiation, Amirkabir University of Technology, Tehran, Iran
3- Associate Professor of Biomedical Engineering, Department of Medical Sciences, Tehran University, Tehran, Iran
4- Assistant Professor of Computer Engineering, Department of Computer Engineering, SR.C., Islamic Azad University, Tehran, Iran
Abstract: (476 Views)
Abstract: Human speech can convey multiple emotions simultaneously; accurately interpreting them is important for understanding the complexity of human communication. Traditional speech emotion recognition systems usually detect only the dominant emotion or rely on multimodal inputs, such as visual and auditory cues. This study aims to use the Brain Emotion Learning model to recognize overlapping and mixed emotional speech. Method: Overlapping emotional speech is synthesized from the standard Berlin Emotional Speech Database using convolution. The Berlin database includes seven emotional states: anger, boredom, disgust, fear, happiness, sadness, and neutral. First, the audio files are processed by MATLAB software. Subsequently, twenty-one overlapping emotional states are generated by convolving two different emotional utterances spoken by the same speaker with identical linguistic content. In the next step, the feature vector for each emotional utterance is extracted. These vectors are then used as input to the brain emotional learning model. Finally, the performance of the proposed model in recognizing overlapping emotions is compared with that of conventional machine-learning models. Results: The findings indicate that although the exact-match recognition accuracy of mixed emotional states is 30% for the brain emotional learning model, the Macro-F1 score of 53% demonstrates that at least one of the emotions present in each overlapping emotional utterance is correctly identified to an acceptable extent. Conclusion: The brain emotional learning model provides a realistic reflection of its performance under real-world conditions when recognizing overlapping emotional speech.
Keywords: Overlapping Emotional Speech, Mixed Emotional Speech, Brain Emotional Learning, Speech Emotion Recognition, Multi-label Emotion Classification
Type of Study: Research |
Received: 2026/08/11 | Accepted: 2026/08/27
Received: 2026/08/11 | Accepted: 2026/08/27
Send email to the article author
| Rights and permissions | |
 |
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License. |
