hzhongresearch

Henry Zhong Research

If you would like to collaborate on research, contact me using: hhcz728 (at) gmail (dot) com .

Research Interests:

Publications

Google Scholar Link

Research Gate Link

Code/Data Samples

AHEAD-DS and OpenYAMNet/YAMNet+

Paper Code Data Model Data2 Website2

Another HEaring AiD scenes Data Set (AHEAD-DS) is a publicly available audio dataset of audiologically relevant environments for hearing aids. OpenYAMNet/YAMNet+ is a lightweight audio recognition model for deployment on resource constrained edge devices, it is a fully open source re-implementation of YAMNet.

The goal of this project was to design a workflow and toolchain to deploy audio recognition models to smartphones connected to wireless earphones. This project offers a baseline audio recognition model and a standardised dataset for comparing audio recognition models. This is a first step to creating AI enhanced hearing aids.

If you use this work please cite our paper.

@article{zhong2025dataset,
  title={A dataset and model for recognition of audiologically relevant environments for hearing aids: AHEAD-DS and YAMNet+},
  author={Zhong, Henry and Buchholz, J{\"o}rg M and Maclaren, Julian and Carlile, Simon and Lyon, Richard},
  journal={arXiv preprint arXiv:2508.10360},
  year={2025}
}

VeinDeep

Paper Code Data

VeinDeep is a system which uses an infrared (IR) depth sensor to extract vein patterns from a person’s hand. The idea is in the future smartphones will be equipped with Kinect V2 like IR depth sensors. Such sensors can be used to identify the smartphone owner using vein patterns and provide a way to unlock the phone. This project was created during my PhD. The associated research paper was presented at PerCom 2017.

If you use this work please cite our paper.

@inproceedings{zhong2017veindeep,
	title={VeinDeep: Smartphone unlock using vein patterns},
	author={Zhong, Henry and Kanhere, Salil S and Chou, Chun Tung},
	booktitle={Pervasive Computing and Communications (PerCom), 2017 IEEE International Conference on},
	pages={2--10},
	year={2017},
	organization={IEEE}
}

QuickFind

Paper Code Data

QuickFind is a fast segmentation and object detection algorithm using only depth maps. Depth maps are images captured from depth sensors like Kinect. The idea is in the future depth sensors will be common so such an algorithm will be useful. This project was created during my PhD. The associated research paper was presented at PerCom Workshops 2016.

The data contains amended ground truth of the RGB-D Scenes dataset used in the QuickFind paper. The ground truth in the original RGB-D scenes dataset had many errors so I manually traced out the location of each object. Credit for RGB-D Scenes goes to Lai et al. Please visit their site and credit them if you use my amended data in conjunction with their RGB-D Scenes dataset.

If you use this work please cite our paper.

@inproceedings{zhong2016quickfind,
	title={Quickfind: Fast and contact-free object detection using a depth sensor},
	author={Zhong, Henry and Kanhere, Salil S and Chou, Chun Tung},
	booktitle={Pervasive Computing and Communication Workshops (PerCom Workshops), 2016 IEEE International Conference on},
	pages={1--6},
	year={2016},
	organization={IEEE}
}

CV

Education

Degrees Schools Dates
Doctor of Philosophy in Computer Science UNSW Sydney 2011 - 2017
Ubiquitous Computing, Computer Vision    
Bachelor of Engineering and Commerce UNSW Sydney 2005 - 2010
1st Class Honours - Software Engineering and Actuarial Studies.    

Experience

Organisations Roles Dates
Macquarie University Postdoc 2024 -
Spiral Blue Co-founder, Head of AI 2018 - 2024
Vardogyir Research and Development 2018 - 2020
UNSW Sydney Research Assistant 2015 - 2017
NEC/UNSW Sydney Research Collaborator 2011 - 2014
NICTA (now CSIRO/Data61) Intern Developer 2009 - 2011
Macquarie Group Intern Developer 2007 - 2008

Skills

Tech - C#, C++, Keras, OpenCV, Pandas, Python, Pytorch, Tensorflow.

Areas of research - Ubiquitous Computing, Computer Vision, Computer Audition