about

Note I am still looking for the job.

Who am I?

Hi, I’m Yang Liu (刘阳)! I am a full-time Microsoft intern researcher since Sept 2019 and work in the Applied Sciences Group led by Dr Vivek Pradeep. My manager is Dr Erci Sommerlade. I am also being a reviewer for IET Single Processing and IET Computer Vision.

I got my Doctor degree at the CVSSP in the University of Surrey at Dec 2019 and passed my VIVA on 1 Nov. 2019 (Examiners are Prof Lyudmila Mihaylova and Dr Jean-Yves Guilemaut). In my PhD study, my supervisor is Prof Wenwu Wang and Prof Adrian Hilton. I also closely collaborated with Prof Jonathon Chambers. My research area is GAN, data fusion, multi-target tracking and optimization in high dimensions.

Education

  • 10/2015 - 11/2019, PhD. in Electric Engineering, University of Surrey, UK.
  • 09/2015 - 03/2015, Master in Electric Engineering and Automation, Harbin Engineering University, China.
  • 09/2008 - 06/2012, Bachelor in Electric Engineering and Automation, Harbin Engineering University, China.

Working Experience

  • 09/2019 - Present Research (intern) ,Microsoft, UK
  • 05/2019 - 05.2019 Volunteer, IEEE International Conference on Acoustics, Speech and Signal Processing, UK
  • 07/2011 - 08/2011 Engineer (intern) ,Siemens , China

Research Project

Audio-Visual Classification and Representation with GAN and VAE

  • 07/2019 - Current | Microsoft, UK
  • Propose a audio scene classification network based on GAN and VAE and sunmit a paper to IJCAI 2020.
  • Train the model on Youtube BB (380,000 videos), Youtube 8M (10,000 videos), Place 365 and Dcase dataset.

S3A: Future Spatial Audio for An Immersive Listener Experience at Home

  • 10/2015 - 06/2019 | Centre for Vision, Speech and Signal Processing, Surrey, UK
  • Work with Prof. Wenwu Wang and Prof. Adrian Hilton and propose a mutli-speaker tracking framework with a microphone array and a camera using DOA, MUSIC, faster R-CNN and YOLO network.
  • The muli-sensor data are fused by sequential Monte Carlo, Probability Hypothesis Density (PHD) filter and particle flow.
  • Implement the methods by Matlab, C++ and Python on AV16.3 (135 videos), AVDIAR (23 videos) and CLEAR (50 videos).
  • Compared to the baseline lines, the tracking filter decrease 52\% accuracy errors and 24\% computational cost.
  • Nominated for Best student paper in International Conference on Acoustics, Speech, and Signal Processing (ICASSP) 2019.

Target Tracking using COZMO robot

  • 07/2018 - 08/2018 | Centre for Vision, Speech and Signal Processing, Surrey, UK
  • Implement a single target tracking for a cozmo robot with YOLO network and Kalman filter by Python and C++.
  • Design a 3D visualizer for Cozmo’s world state by Qt, OpenCV and OpenGL.

IEEE-AASP Challenge on Acoustic Source Localization and Tracking

  • 06/2018 - 07/2018 | Centre for Vision, Speech and Signal Processing, Surrey, UK
  • Propose a mutli-speaker tracking framework based on particle flow with MUSIC method.
  • The method is evaluated on the LOCATA datasets which are recorded by four microphone arrays (20 recordings).
  • Compared to the baseline method, our proposed method increases 42\% the probability of detection and decreases 71\% the elevation error on the planar microphone array.

Teaching

  • Computer Algorithms and Architecture (80 student) and Computers and Programming 2 (80 student) about C++ with Dr. Jean-Yves Guillemaut. (Jan 2017 – Oct 2018, 1 yr 10 mos)
  • Al and Al Programming (40 students) and Advanced Signal Processing (40 students) about Matlab and machine learning with Dr. Terry Windeatt. (Jan 2017 – May 2017, 5 mos)
  • Web and Database Systems (70 students) about SQL, PHP and HTML with Prof. Shujun Li. (Oct 2016 – Feb 2017, 5 mos)
  • Computer and Digital Logic (80 students) about Python with Dr. Nikolaos Dikaios. (Sep 2016 – Dec 2016, 4 mos)

Publication

Journal

  • Labelled non-zero particle flow for Multi-speaker tracking, Yang Liu, Wenwu Wang, IEEE Transactions on Signal processing (under review).
  • Intensity Particle Flows for Sequential Monte Carlo Implementation of Probability Hypothesis Density Filter, Yang Liu, Wenwu Wang, IEEE Transactions on Signal processing (under review).
  • Audio-visual Zero Diffusion Particle Flow SMC-PHD Filter for Multi-speaker Tracking, Yang Liu, Volkan Kili̧c, Jian Guan, Wenwu Wang, IEEE Transactions on Multimedia, August 2019.
  • Texture features extraction method based on Worldview-II multi spectral remote sensing data, Zhenxing Zhang, Ning Li, Yang Liu, Systems Engineering and Electronics, 2013, 35(10): 2044-2049.

    Conference

  • Labelled Non-zero Particle flow for SMC-PHD filtering, Yang Liu, Qinghua Hu, Yuexian Zou, Wenwu Wang, International Conference on Acoustics, Speech, and Signal Processing, 2019.
  • Intensity Particle Flow SMC-PHD Filter For Audio Speaker Tracking, Yang Liu, Wenwu Wang, Volkan Kılıc, LOCATA challenge workshop, 2018.
  • Audio-visual SMC-PHD Filter with Non Zero Diffusion Particle Flow, Yang Liu, Wenwu Wang, Volkan Kılıc, International Conference on Acoustics, Speech, and Signal Processing, 2018.
  • Particle flow for sequential Monte Carlo implementation of probability hypothesis density, Yang Liu, Wenwu Wang, Yuxin Zhao, International Conference on Acoustics, Speech, and Signal Processing, 2017.
  • Particle Flow SMC-PHD Filter for Audio-Visual Multi-speaker Tracking, Yang Liu, Wenwu Wang, onathon Chambers, Adrian Hilton, International Conference on Latent Variable Analysis and Signal Separation, 2017.
  • Visual Mapping and Localization Using a Tree-structured Audio Model, Yuxin Zhao, Yang Liu, Wenwu Wang, International Navigation Conference, 2015.