See Proceedings in ISCA archive
Tuesday 23rd
- 08:00 – 09:00Registration
- 09:00 – 09:30Opening
- 09:30 – 10:30Keynote 1: Rigorous Forensic Automatic Speaker Recognition: Bayesian Decision Theory, Probabilistic Calibration and Case-Specific Validation – Daniel Ramos (Chair: Alicia Lozano-Diez) – Keynote Presentation
- 10:30 – 11:00Coffee Break
-
11:00 – 12:20Oral Presentations 1.1 – Deepfake and Spoofing Detection
Session Chair: Kong Aik Lee, The Hong Kong Polytechnic University- 11:00 An Intervention-Based Framework for Shortcut Diagnosis in Spoofing Countermeasures
- 11:20 Domain Adaptation for Deepfake Audio Detection under Degraded Channel Conditions
- 11:40 Speaker-Invariant Representation Learning for Spoofing Detection via Gradient Reversal and Information Bottleneck
- 12:00 Large-Kernel 1D CNN for Raw Waveform Spoofing Countermeasures
- 12:20 – 13:45Lunch
-
13:45 – 15:30Early-Stage Researcher Symposium (ESRS) – Poster Session
Session Chair: Tom Bäckström, Aalto University- Why Do You Say It Like That? A Phoneme-Level Framework for Explainable Speech Deepfake Detection
- Interpreting SSL Representations for Spoof Detection: a WavLM Study
- Identity Disambiguation in Common Voice: Enabling Fairness Evaluation Across Demographic Subgroups
- Machine-Learning Benchmarking of Voice-Based Biomarkers for Parkinson’s Disease
- The Role of Voice Source and Filter in Speech Emotion Recognition
- Transparent Exchange of Speaker Attributes
- Limitations of WER for Intelligibility Evaluation in Speech Anonymization
- Controllable Voice Anonymization for Privacy-Preserving Disease Detection from Speech
- Challenges in Multi-Speaker Privacy
- Studying Voice Privacy Risks with Side Information through Partially Synthetic Data
- Challenges in Protection against Deepfakes in Speech
- Linkage-Based Adversarial Framework for Voice Privacy Evaluation
- Why Voice Privacy Researchers Should Worry About Attribute Inference?
- How Bilingual Are SSL Speech Models? Cross-Lingual Probing of Articulatory Encoding with Finnish and Russian EMA
- 15:30 – 16:00Coffee Break
-
16:00 – 17:00Oral Presentations 1.2 – Privacy-Aware Speech Processing and Watermarking
Session Chair: Lauri Juvela, Aalto University -
17:00 – 18:00Oral Presentations 1.3 – Tools and Methods for Speaker Verification
Session Chair: David Martínez González, Hiya Inc. - 18:30 – 20:00Welcome Reception
Wednesday 24th
-
08:30 – 09:30Oral Presentations 2.1 – Speaker Diarization
Session Chair: Alfonso Ortega, Universidad de Zaragoza - 09:30 – 10:30Keynote 2: Genetic information in human voice: how much do we know today and how much more will technology uncover? – Rita Singh (Chair: Javier Hernando) – Keynote presentation
- 10:30 – 11:00Coffee Break
-
11:00 – 12:20Oral Presentations 2.2 – Privacy, Security and Anonymization
Session Chair: Jean-François Bonastre, AMIAD & Université d’Avignon - 12:20 – 13:45Lunch
-
13:45 – 15:30Special Sessions – Oral Overviews (15 min. each) followed by Poster presentations
- 13:45 [SS1] Special Session on Speech and Language Technologies in Healthcare – Oral Overview
- 14:00 [SS2] Special Session on Model Fairness Meets Source Tracing: Toward Trustworthy AI for Manipulated Speech Attribution – Oral Overview
- 14:15 [SS3] Special Session on NIST SRE24 Deeper Analysis – Oral Overview
- 14:30 [SS4] Special Session on TidyLang Challenge: Speaker-Controlled Language Recognition – Oral Overview
- 14:45 – 15:30 Poster presentations:
- [SS1] MEGConformer: Conformer-Based MEG Decoder for Robust Speech and Phoneme Classification
- [SS1] Adaptive Phone-Wise Weighted Loss for Silent Speech Restoration in Continuous Spanish
- [SS1] Comparator Loss: An Ordinal Contrastive Loss to Derive a Severity Score for Speech-based Health Monitoring
- [SS1] Rapid Calibration for Cross-Subject Imagined Speech Decoding Toward Restoring Communication
- [SS1] Deep learning based analysis of spontaneous speech for diagnostic classification and biomarker prediction in Alzheimer’s disease and primary progressive aphasia
- [SS1] Vocal markers of Turner syndrome: a preliminary analysis of sustained vowel recordings
- [SS1] End-to-End Intracortical Speech Decoding from Neural Activity
- [SS1] FiLM-Based Speaker Conditioning of a SpeechLLM for Pathological Speech Recognition
- [SS2] The Effect of Telephony Transmission on Source Tracing of Audio Deepfakes
- [SS2] Advancing Zero-Shot Open-Set Speech Deepfake Source Tracing
- [SS3] I4U’s Official and Streamlined Audio Systems for NIST SRE24
- [SS3] Analysis of the NIST 2024 Speaker Recognition Evaluation
- [SS4] Spoken Language Identification with Pre-trained Models and Margin Loss
- [SS4] LLM-Based Language Verification and Multimodal Ensemble for Spoken Language Recognition
- [SS4] Speaker-Aware Language Verification Based on Attentive Pooling, Mixture of Experts and Neural PLDA
- 15:30 – 16:00Coffee Break
-
16:00 – 17:00Oral Presentations 2.3 – Optimization and Efficiency in Speaker Recognition
Session Chair: Tanel Alumäe, Tallinn University of Technology- 16:00 Harmonizing data augmentation and loss function for speaker recognition: examples with speed perturbation, mixup and mixout
- 16:20 Assessing the Energy and Carbon Emissions of Neural Speaker Verification Model in Training and Inference
- 16:40 On Low-Bit Quantization Errors in Speaker Verification: Diagnostic and Mitigation
-
17:00 – 18:00Oral Presentations 2.4 – Speech biomarkers and Speech quality
Session Chair: Eva Navas, University of the Basque Country- 17:00 SLAP: Learning Speaker and Health-Related Representations from Natural Language Supervision
- 17:20 Speech Quality Embeddings for Improved Detection and Classification of Degradations in Speech Signals
- 17:40 Dysarthria Severity Classification on the HeyJay! Dataset: A Parameter-Efficient Approach Using Self-Supervised Speech Representations
- 19:30 – 23:30Gala Dinner
Thursday 25th
-
08:30 – 09:30Oral Presentations 3.1 – SSL Models for Spoofing Detection
Session Chair: Massimiliano Todisco, EURECOM- 08:30 A comparison of SSL-Based Feature Extractors and Back-End Classifiers for Spoofing Detection: A Multi-Corpus Training and Cross-Linguistic Analysis
- 08:50 From Self-Supervised Speech Models to Mixture-of-Experts for Robust Anti-Spoofing
- 09:10 Can SSL Frontend Generalize to All-Type Audio Spoofing?
- 09:30 – 10:30Keynote 3: Every breath you take: From Vocal Chords to Health Scores – Björn Schuller (Chair: José A. González-López) – Keynote presentation
- 10:30 – 11:00Coffee Break
-
11:00 – 12:20Oral Presentations 3.2 – Backend and Generalization in Speaker Verification
Session Chair: Douglas Reynolds, MIT Lincoln Lab- 11:00 Spherical-Gaussian TPSDA: combining PLDA, T-PSDA and duration models for speaker verification
- 11:20 Condition-Aware System Fusion for Speaker Verification
- 11:40 Towards Language-Agnostic Speaker Verification: A Cross-Lingual Transfer Study of Architectures
- 12:00 Subtract to Clean, Add to Enrich: Dual-Path Disentanglement for Speaker and Language Recognition
- 12:20 – 13:45Lunch
- 14:00 – 18:30/19:00Tour to Cascais and Sintra
Friday 26th
-
08:30 – 09:30Oral Presentations 4.1 – Representation Learning in Speaker and Language
Session Chair: Jonas Borgstrom, MIT Lincoln Laboratory - 09:30 – 10:30Keynote 4: From Single-Channel Foundations to Multi-Speaker and Multi-Modal Understanding – Lukáš Burget (Chair: Isabel Trancoso) – Keynote Presentation
- 10:30 – 11:00Coffee Break
-
11:00 – 12:20Oral Presentations 4.2 – Spoofing Detection and Robust ASV
Session Chair: Elie Khoury, Pindrop- 11:00 Sparse deepfake detection promotes better disentanglement
- 11:20 I Hear, Therefore I Trust: A Socio-Technical Investigation of Humans as Synthetic Speech Detectors
- 11:40 PLDA Scoring for Spoofing-Robust Automatic Speaker Verification
- 12:00 J-SPAW2: A Japanese Corpus for Speaker Verification and Anti-Spoofing with Challenging Replay and Speech Synthesis Attacks
- 12:20 – 13:00Closing Ceremony