Undergraduate Researcher in Machine Learning & Audio Intelligence
BSc Information and Computing Science · Xi'an Jiaotong-Liverpool University (XJTLU)
I am interested in building machine learning systems for audio understanding, speech processing, and data-driven intelligent applications. My research interests include acoustic scene analysis, sound event detection, speech enhancement, domain adaptation, and representation learning.
Audio Machine Learning · Acoustic Scene Analysis · Sound Event Detection · Speech Enhancement · Domain Adaptation · Deep Learning
Research implementation developed during the XJTLU SURF 2025 project Data-Efficient and Low-Complexity Acoustic Scene Classification.
The project investigates device and city domain shifts in acoustic scene classification using a shared time-frequency representation, dual scene-classification heads, and domain-adversarial learning.
Highlights: domain adaptation · time-frequency modeling · dual-head classification · PyTorch
Previous research work based on PrimeK-Net for single-channel speech enhancement, including strict baseline reproduction and subsequent phase-oriented analysis.
The study examined phase concentration and statistical phase modeling as possible directions for improving phase-aware speech enhancement.
Topics: speech enhancement · baseline reproduction · spectral learning · phase analysis
Current research preparation for my Final Year Project focuses on audio detection and sound event detection, including baseline systems, pretrained audio representations, ATST-based approaches, and two-stage training strategies.
The goal is to identify a well-grounded model-improvement direction while maintaining compatibility with established detection baselines.
A mobile-first web experience for exploring the cultural heritage of Maple Bridge, combining interactive maps, route guidance, multimedia content, accessibility-oriented interactions, and a lightweight local FAQ assistant.
Tech: HTML · CSS · JavaScript · AMap Web SDK · Responsive Design
A 9-member team project for CPT202 developing a cultural heritage content platform with role-based publishing, review, and community workflows.
My contribution focused on the backend administrative workflow, including article moderation, content management, user and contributor permission flows, JWT authentication with Spring Security, and API-level regression testing.
Tech: Java · Spring Boot · Spring Security · JWT · Spring Data JPA · REST APIs · MockMvc
View final team repository →
View my contribution snapshot →
Programming: Python · Java · C++ · JavaScript
Machine Learning: PyTorch · deep learning · audio ML · model evaluation
Software Engineering: Git · REST APIs · frontend/backend development · testing
Research: experiment design · literature review · reproducibility · technical documentation
Xi'an Jiaotong-Liverpool University
BSc Information and Computing Science
I am currently working on:
- audio detection and sound event detection for my Final Year Project;
- pretrained audio representations and two-stage training strategies;
- identifying reproducible model-improvement directions compatible with established sound event detection baselines.
GitHub: @AlbertXia-herald