Research Demo

My research focuses on audio and visual signal processing, robotic perception and machine learning. I developed microphone array techniques for sound enhancement, source localizaiton and blind source separtation. I developed audio-visual signal processing techniques for acoustic sensing from flying robots (mini-drones). I applied machine learning techniques to human activity and context recogntiion from wearable sensors (motion, GPS, sound and image). This page shows demos of my research outcomes.

Robotic perception

• TFBSS framework for ego-noise reduction
  The demo presents the state-of-the-art of ego-noise reduction performance achieved by our TFBSS framework.
  
• Embedded multichannel sound acquisition system for drone audition
  The demo presents our latest hardware prototype for drone audition.
  
• Audio-visual sensing dataset and baseline performance
  The demo presents the baseline acoustic sensing results using AVQ, the first audio-visual dataset recorded outdoors from a quadcopter.
  
• Multimodal audio-visual sensing for auditory MAV
  This demo combines audio and visual modalities to localize and enhance multiple speakers from a multirotor MAV platform.
  
• Microphone-array ego-noise reduction for audiotory MAV
  The demo compares state-of-the-art microphone-array algorithms for ego-noise reduction and sound enhancement from a multirotor MAV platform.
  

Microphone array

• Pseudo-determined BSS for ad-hoc microphone networks
  This demos performs blind source separation from crowsourced audio recording in a cocktail party environment (with 8 microphones and 4 speakers).
  
• BSS + beamforming
  The demos extracts a target speaker in an extremely noisy cocktail party environments (SNR < -10 dB) by combining beamforming and blind source separation.
  
• BSS-MBMC
  This demos shows blind source separation results (with 4 microphones and 4 speakers) obtained by the MBMC permutation ambiguity alignment algorithm.
  

Machine learning

• SHL Dataset
  
• SHL challenge