Synthetic Sound Detection
Distinguishing real gunshots and explosions from AI-generated sounds using learned audio representations and phase-based features.
Environmental audio · Group delay
UNIVERSITY AT BUFFALO / MDFL
Understanding sound.
Building trust.
We study the signals behind what we hear—from the authenticity of digital audio to the human stories carried in speech.
Audio Team · UB Media Forensics Lab
Explore our research
AUDIO INTELLIGENCEBUFFALO, NY
01 / RESEARCH
Listening beyond
the surface.
We connect signal processing and machine learning to understand audio, trace its origins, and make speech technology more useful.
01
Audio Security & Forensics
Understanding what is real, what changed, and how.
Cross-Codec Speech Deepfake Detection
Studying how detectors transfer across neural codecs, and selecting complementary codecs for more efficient training.
Neural codecs · Transfer-aware selection
Partial Speech Deepfake Localization
Locating manipulated segments within speech and examining how detection and localization change across datasets.
Partial deepfakes · Cross-dataset evaluation
Audio Provenance Reconstruction
Reconstructing which transformations a recording has undergone—and the order in which they were applied.
Audio transformations · Sequence modeling
02
Audio Watermarking
Following a signal through change.
Watermarks that withstand transformation
Embedding and recovering audio watermarks under compression, trimming, and voice cloning. We investigate redundant embedding and the behavior of signal-level and identity-related watermark approaches.
Watermark recovery · Voice cloning · Redundancy
03
Healthcare & Speech
Making audio intelligence work for people.
Cardio-Respiratory Audio Analysis
Developing hierarchical, multitask models to analyze heart sounds, breathing, and coughs for disease classification research.
Heart & lung sounds · Multitask learning
Children’s Speech Normalization
Investigating speech recognition and normalization for children, with a focus on speech from children with developmental disabilities.
Children’s speech · Recognition & normalization
02 / OUR PEOPLE
The team behind
the signal.
A shared curiosity about sound.
A collaborative approach to research.
Professor / Director
Siwei Lyu
siweilyu@buffalo.eduAssistant Director
Chengzhe Sun
csun22@buffalo.eduAudio Team Technical Leader
JaeHyeong Chang
jchang46@buffalo.edu03 / PUBLICATIONS
Research, shared.
PUBLICATION UPDATES
Our publication list is on its way.
Paper details, links, and accompanying research resources will be added here as they become available.
04 / GET IN TOUCH
Let’s talk
audio research.
For research questions and collaboration inquiries, contact our Audio Team Technical Leader.
jchang46@buffalo.eduJaeHyeong Chang
University at Buffalo, SUNY