publications

2027

1. CSL

   Compositional Domain Adaptation for Automatic Speech Recognition with Headwise Selective Attention Merging

   Natarajan Balaji Shankar, Z. Wang, Eray Eren, and 1 more author

   Computer Speech & Language, 2027

   Abs DOI Bib

   

   Extends headwise attention-merging techniques to compositional domain adaptation for automatic speech recognition, combining models trained on different domains at the attention-head level without joint retraining.

   @article{shankar2027compositional,
     title = {Compositional Domain Adaptation for Automatic Speech Recognition with Headwise Selective Attention Merging},
     author = {Shankar, Natarajan Balaji and Wang, Z. and Eren, Eray and Alwan, Abeer},
     journal = {Computer Speech & Language},
     volume = {101},
     pages = {102012},
     year = {2027},
     doi = {10.1016/j.csl.2026.102012},
     url = {https://scholar.google.com/citations?view_op=view_citation&hl=en&user=cjhgbFEAAAAJ&citation_for_view=cjhgbFEAAAAJ:Tyk-4Ss8FVUC}
   }

2026

1. JASA-EL

   Improving Zero-Shot Style Transfer Text-to-Speech by Disentangled Fine-Grained Style Modeling

   Eray Eren, Q. Liu, Abeer Alwan, and 1 more author

   JASA Express Letters, 2026

   Abs Bib

   

   A lightweight (48M-parameter) zero-shot style-transfer text-to-speech system with fine-grained style disentanglement, achieving a 3.64 prosody pMOS score and significantly outperforming competitive baselines.

   @article{eren2026styletransfer,
     title = {Improving Zero-Shot Style Transfer Text-to-Speech by Disentangled Fine-Grained Style Modeling},
     author = {Eren, Eray and Liu, Q. and Alwan, Abeer and Bharaj, Gaurav},
     journal = {JASA Express Letters},
     volume = {6},
     number = {3},
     pages = {034802},
     year = {2026},
     url = {https://scholar.google.com/citations?view_op=view_citation&hl=en&user=cjhgbFEAAAAJ&citation_for_view=cjhgbFEAAAAJ:zYLM7Y9cAGgC}
   }

2. ICASSP

   STACodec: Semantic Token Assignment for Balancing Acoustic Fidelity and Semantic Information in Audio Codecs

   K. Zhang, M. Shi, Eray Eren, and 3 more authors

   In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2026

   Accepted

   Abs arXiv Bib

   

   Introduces semantic token assignment to balance acoustic fidelity and semantic information in neural audio codecs, aiming to improve downstream speech-language-model performance without sacrificing reconstruction quality.

   @inproceedings{zhang2026stacodec,
     title = {STACodec: Semantic Token Assignment for Balancing Acoustic Fidelity and Semantic Information in Audio Codecs},
     author = {Zhang, K. and Shi, M. and Eren, Eray and Shankar, Natarajan Balaji and Wang, Z. and Alwan, Abeer},
     booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
     year = {2026},
     note = {Accepted},
     url = {https://arxiv.org/abs/2602.06180}
   }

3. INTERSPEECH

   Entropy-Aware Domain-Routed Mixture-of-Experts Speech-LLM Framework: A Case Study of Multi-Domain Child-Adult ASR

   M. Shi, K. Zhang, Z. Wang, and 3 more authors

   In Proc. INTERSPEECH, 2026

   Accepted

   Abs arXiv Bib

   

   Proposes an entropy-aware routing mechanism for a domain-routed mixture-of-experts speech-LLM architecture, assigning experts based on input uncertainty across speech domains.

   @inproceedings{shi2026entropy,
     title = {Entropy-Aware Domain-Routed Mixture-of-Experts Speech-LLM Framework: A Case Study of Multi-Domain Child-Adult ASR},
     author = {Shi, M. and Zhang, K. and Wang, Z. and Shankar, Natarajan Balaji and Eren, Eray and Alwan, Abeer},
     booktitle = {Proc. INTERSPEECH},
     year = {2026},
     note = {Accepted},
     url = {https://arxiv.org/abs/2606.10454}
   }

2025

1. INTERSPEECH

   ProMode: A Speech Prosody Model Conditioned on Acoustic and Textual Inputs

   Eray Eren, Q. Liu, H. Kim, and 2 more authors

   In Proc. INTERSPEECH, 2025

   Abs arXiv Bib

   

   A standalone zero-shot speech prosody model using a Perceiver-IO architecture to map text and reference audio to latent prosodic features (pitch, duration, and energy), achieving state-of-the-art prosody-prediction accuracy.

   @inproceedings{eren2025promode,
     title = {ProMode: A Speech Prosody Model Conditioned on Acoustic and Textual Inputs},
     author = {Eren, Eray and Liu, Q. and Kim, H. and Garrido, P. and Alwan, Abeer},
     booktitle = {Proc. INTERSPEECH},
     year = {2025},
     url = {https://arxiv.org/abs/2508.09389}
   }

2. ICASSP

   Selective Attention Merging for Low Resource Tasks: A Case Study of Child ASR

   Natarajan Balaji Shankar, Z. Wang, Eray Eren, and 1 more author

   In IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2025

   Abs Bib

   

   Applies selective, headwise attention-merging between models to adapt automatic speech recognition to the low-resource domain of children’s speech.

   @inproceedings{shankar2025selective,
     title = {Selective Attention Merging for Low Resource Tasks: A Case Study of Child ASR},
     author = {Shankar, Natarajan Balaji and Wang, Z. and Eren, Eray and Alwan, Abeer},
     booktitle = {IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
     year = {2025},
     url = {https://scholar.google.com/citations?view_op=view_citation&hl=en&user=cjhgbFEAAAAJ&citation_for_view=cjhgbFEAAAAJ:2osOgNQ5qMEC}
   }

2023

1. CSL

   Deep Learning-Based Speaker-Adaptive Postfiltering with Limited Adaptation Data for Embedded Text-to-Speech Synthesis Systems

   Eray Eren and Cenk Demiroglu

   Computer Speech & Language, 2023

   Abs DOI Bib

   

   Proposes few-shot, speaker-adaptive neural postfilters (including Transformer- and GAN-based architectures) for data- and parameter-efficient embedded text-to-speech synthesis.

   @article{eren2023postfiltering,
     title = {Deep Learning-Based Speaker-Adaptive Postfiltering with Limited Adaptation Data for Embedded Text-to-Speech Synthesis Systems},
     author = {Eren, Eray and Demiroglu, Cenk},
     journal = {Computer Speech & Language},
     volume = {81},
     pages = {101520},
     year = {2023},
     doi = {10.1016/j.csl.2023.101520},
     url = {https://doi.org/10.1016/j.csl.2023.101520}
   }

2. INTERSPEECH

   FusedF0: Improving DNN-Based F0 Estimation by Fusion of Summary-Correlograms and Raw Waveform Representations of Speech Signals

   Eray Eren, Lee Ngee Tan, and Abeer Alwan

   In Proc. INTERSPEECH, 2023

   Abs Bib

   

   A hybrid DSP and DNN F0 estimation architecture that fuses noise-robust summary correlograms with raw waveform representations for robust pitch extraction.

   @inproceedings{eren2023fusedf0,
     title = {FusedF0: Improving DNN-Based F0 Estimation by Fusion of Summary-Correlograms and Raw Waveform Representations of Speech Signals},
     author = {Eren, Eray and Tan, Lee Ngee and Alwan, Abeer},
     booktitle = {Proc. INTERSPEECH},
     pages = {4523--4527},
     year = {2023},
     url = {https://scholar.google.com/citations?view_op=view_citation&hl=en&user=cjhgbFEAAAAJ&citation_for_view=cjhgbFEAAAAJ:9yKSN-GCB0IC}
   }

2022

1. SpeechCom

   Uncertainty Assessment for Detection of Spoofing Attacks to Speaker Verification Systems Using a Bayesian Approach

   Cagil Suslu, Eray Eren, and Cenk Demiroglu

   Speech Communication, 2022

   Abs DOI Bib

   

   Uses a Bayesian neural network approach to estimate predictive uncertainty for detecting spoofing attacks against automatic speaker verification systems.

   @article{suslu2022uncertainty,
     title = {Uncertainty Assessment for Detection of Spoofing Attacks to Speaker Verification Systems Using a Bayesian Approach},
     author = {Suslu, Cagil and Eren, Eray and Demiroglu, Cenk},
     journal = {Speech Communication},
     volume = {137},
     pages = {44--51},
     year = {2022},
     doi = {10.1016/j.specom.2021.12.003},
     url = {https://doi.org/10.1016/j.specom.2021.12.003}
   }

2021

1. Thesis

   Deep Learning-Based Speaker-Adaptive Postfiltering with Limited Adaptation Data for Embedded Text-to-Speech Synthesis Systems

   Eray Eren

   Ozyegin University, 2021

   M.S. thesis, advised by Prof. Cenk Demiroglu

   Abs Bib

   

   M.S. thesis on few-shot, speaker-adaptive neural postfilters (including Transformer- and GAN-based approaches) for data- and parameter-efficient text-to-speech synthesis on embedded systems.

   @mastersthesis{eren2021thesis,
     title = {Deep Learning-Based Speaker-Adaptive Postfiltering with Limited Adaptation Data for Embedded Text-to-Speech Synthesis Systems},
     author = {Eren, Eray},
     school = {Ozyegin University},
     address = {Istanbul, Turkey},
     year = {2021},
     note = {M.S. thesis, advised by Prof. Cenk Demiroglu},
     url = {https://scholar.google.com/citations?view_op=view_citation&hl=en&user=cjhgbFEAAAAJ&citation_for_view=cjhgbFEAAAAJ:u-x6o8ySG0sC}
   }