Hzpsbxvavmwlqysved0f

Wen-Chuan Chen

Deep Learning Researcher  •  Hsinchu, TW  •  [email protected]

Tech enthusiast. Addicted to innovative products and arming myself with state-of-the-art research. Major in deep learning applied to digital audio/image processing, computer vision, speech recognition, and self-driving car.

Education

National Tsing Hua University, Master of Science (MS), Power Mechanical Engineering, 2017 ~ 2019

  • Telecom Electro-Acoustic Audio Lab directed by Dr. Mingsian R. Bai
  • GPA: 4.05

University@2x

National Sun Yat-sen university, Bachelor of Science (BS), Mechanical and Electro-mechanical Engineering, 2013 ~ 2017

University@2x

Skills


Programming skills

  • Tensorflow / Keras / Pytorch
  • Python - OO design and multithread programming
  • C / C++ - Data structure and algorithms
  • Matlab - Scientific computing and algorithms design


Language

  • Chinese - Native
  • English - TOEIC 920

Publication

Master's Thesis: "Deep Learning Applied to Speech and Audio Signal Processing under Adverse Environments", National Tsing Hua University, AUG. 01, 2019 in Taiwan.

The thesis discuss the real-life tasks such as dereverberation, sound event detection, and source separation under adverse environments, especially highly reverberant. Traditional Algorithms are compared with proposed deep neural network based methods. The results show that proposed methods outperform others.

Journal Paper: "Multichannel Sound Event Separation and Detection in Adverse Environments Using Convolutional Recurrent Neural Network with Complex Masking," IEEE/ACM Transactions on Audio, Speech and Language Processing (Under Review).

Propose an end-to-end polyphonic sound event separation and detection system aimed at separating and detecting multiple sound events. The system consists of a convolutional recurrent neural network in conjunction with an ideal complex mask. In addition, a discriminative training network is employed to increase the robustness.

Conference Paper: "Deep Learning Applied to Dereverberation and Sound Event Classification," International Congress on Acoustics, SEPT. 09-13, 2019 in Germany.

Investigates dereverberation and sound event detection techniques, with the aid of deep learning. The proposed system consists of two units: a multichannel deep neural front-end and a VGG-like classifier back-end trained with generic data augmented by various room impulse responses.

Conference Paper: "Robotic Voice Assistant Equipped with Binaural Audio," INTERNOISE, JUNE. 16-19, 2019 in Madrid.

The robot is comprised primarily of three functional units: (1) a microphone array used to locate the user position (2) a cloud-based DNN-based classifier for command words recognition (3) loudspeaker array for binaural audio sound effect.

Honors


  • Attended INTERNOISE conference, JUNE. 16-19, 2019 in Madrid
  • Graduate representative of College of Engineering, NTHU, 2019
  • Industry-academia Collaboration - Smart Acoustic Homecare System, 2017 - 2019
  • Director of Senior Project Contest - Command Word Recognition under Adverse Environments, 2018
  • Organized Chinese Society of Sound and Vibration 2018 at NTHU
  • Activities Director of Guitar club, NSYSU, 2014 - 2015
  • Vocalist Guitar club, NSYSU, 2013 - 2015
  • Director of Public Relations of MEM, NSYSU, 2013 - 2015

Professional Curriculum


  • Machine Learning, Jen-Tzung Chien (A+)
  • Deep Learning, Jen-Tzung Chien (A+)
  • Computer Vision, Min Sun (A)
  • Convolutional Neural Networks for Visual Recognition, Fei-Fei Li (Online Course)
  • Music Information Retrieval, Li Su (A)
  • Digital Control System, Ting-Jen Yeh (A+)
  • Digital Signal Processing, Ming-Sian Bai (A+)
  • Acoustic Array Signal Processing, Ming-Sian Bai (A)
  • Mobile Robots and Self-driving Cars, Ting-Jen Yeh (A)

Projects

Paragraph image 01 00@2x

Source Separation

Source separation, blind signal separation (BSS) or blind source separation, is the separation of a set of source signals from a set of mixed signals, without the aid of information (or with very little information) about the source signals or the mixing process.

 My implementation on Github

Paragraph image 01 00@2x

Panorama Stitching 

Automatic panorama stitching technology has been widely adopted in many applications such as Google Street View, panorama photos on smartphones, and stitching software such as Photosynth and AutoStitch.

 My implementation on Github

Powered by CakeResumePowered by CakeResume
Powered by CakeResumePowered by CakeResume