Research

Graduate stage (Principle investigator: Tao Zhang, Ph.D.) at Tianjin University, Tianjin, China

Efficient convolution neural network (CNN) for mobile applications

Research on the reduction of computational complexity and resource cost without much performance degradation.
Investigation on how to reduce the feature redundancy in CNN architecture.
Extracted feature representions to solve multimedia tasks, including acoustic scene classification, sound event detection and image classification.

Detection and classification of acoustic scenes and events

To recognize an audio scene or a sound event either from a recording or an on-line stream through pattern recognition and signal processing.
Illustrated the relationship between receptive field in CNN and time-frequency feature resolution in mel energy spectrogram.
Proposed a deep CNN architecture using the fine resolution feature.

Image recognition with system on chip (SoC)

To classify the fruit categories in the captured image with TI AM5708.
Responsible for the setup of deep learning environment, models comparison and selection, model compression and optimization.
Successfully classified 98 categories of fruits in a short duration.

Undergraduate stage at Tianjin University, Tianjin, China

Image denoising based on residual learning and batch normalization

To recover images from Gaussian noise with unknown noise level.
Mainly responsible for implementing various denoising methods for comparison.
Experimentally verified the effect of residual learning strategy and batch normalization for CNN denoising.

Research and development on 2-D sonar equipment based on beamforming

To measure the distance and the depth of streambed ahead in real-time.
Developed a suitable beamforming algorithm under the limitation of fixed arrays.
Designed an underwater 2-D detection system using the vernier method.