Qun Feng Tan's CV

Check out a robot Rubik's Cube solver I built:

Google Scholar Profile

Education

• University of Southern California
  Aug 2008 – May 2013 (defended March 2013)
  Ph.D. in Electrical Engineering
  GPA: 4.00/4.00
• Stanford University
  Sept 2006 – June 2008
  M.S. in Electrical Engineering
  GPA: 3.93/4.00
• University of Michigan, Ann Arbor
  Sept 2004 – June 2006
  Entered with 28 transfer credits from Cambridge A Level
  B.S.E in Electrical Engineering and minor in Mathematics (Summa Cum Laude)
  GPA: 3.93/4.00

Academic Emphasis: Machine Learning, Pattern Recognition, Large Datasets, Speech processing, Natural Language Processing, Information Theory, Convex Optimization, Estimation Theory, Probability Theory, Compressive Sensing, and many other things...

Programming Skills

• Programming Language Usage: C++, C, C#, Scripting languages, Java, HTML/CSS
• Speech Related: Moses (Machine Translation), HTK (Automatic Speech Recognition), SRILM, OpenFST

Languages Spoken/Writing/Reading: English, Chinese (Simplified and Traditional), Malay

Languages Spoken: Hokkien, Taiwanese, Hakka, Indonesian

Teaching Assistant experience
Teaching Assistant Evaluation Results (Students' feedback and ratings) for EE 483 (Digital Signal Processing) Spring 2012 Semester.

Rating is on scale from 1 to 5 (5 being the best score).

Obtained perfect (5 out of 5) overall ratings.

Selected Publications (as first author)
Peer Reviewed Journals [Google Scholar Profile]

1. “Novel Variations of Group Sparse Regularization Techniques with Applications to Noise Robust Automatic Speech Recognition” – IEEE Transactions on Audio, Speech, and Language Processing, vol. 20, no. 4, pp. 1337 –1346, May 2012 [preprint]
2. “Enhanced Sparse Imputation Techniques for a Robust Speech Recognition Front-End” – IEEE Transactions on Audio, Speech and Language Processing, vol. 19, no. 8, pp. 2418 – 2429, 2011 [preprint]

Peer Reviewed Conferences

1. “Combining Window Predictions Efficiently - A New Imputation Approach For Noise Robust Automatic Speech Recognition” – ICASSP 2013 [preprint]
2. “Automatic Speech Recognition System Channel Modeling” – Interspeech 2010 [preprint]

• Student member of IEEE (2012)
• Reviewer for Symposium on Machine Learning in Speech and Language Processing, IEEE Transactions on Signal Processing (2012)

Other Interests

➨ Click for method to solve the Rubiks cube really fast

Non-Sighted Rubik Solving Videos

Video 1: Solving the Rubiks Cube behind my back

Solving the cube behind my back in 3 min 30 seconds (time including memorization). Memorization took 60 seconds. (Recorded March 2006)

1) My friend scrambles the cube
2) Memorization takes place (60 seconds - Not filmed since the Canon camera at that time has a video time limit of 3 min)
3) Solving behind my back (2 min 30 seconds)
4) Done

Video 2: Blindfold Solving Video

Solving the cube blindfolded in 2 minutes 37 seconds (time including memorization). Memorization took 52 seconds. (Recorded Dec 11, 2005)

1) I scramble the cube on spot
2) Memorization takes place (52 seconds)
3) Donning the blindfold
4) Solving while blindfolded
5) Done

Some personal Rubiks Cube records....(Circa 2006)

Scrambles are generated using computer program.

1) Speedsolve records:

i) Skipping 2 steps: 7 seconds
ii) Non-lucky: 11 seconds

2) Non-sighted records:

i) Blindfolded (including time for memorization): 2 minutes 06.78 seconds
ii) Behind my back (including time for memorization): 3 minutes 26 seconds
iii) 2 cubes consecutively blindfolded (first is memorizing 2 cubes together, then solving them while blindfolded): 9 minutes 59.28 seconds
iv) Memorize a cube that is randomly scrambled: 45 seconds
v) 4X4X4 cube (Rubik's Revenge) blindfolded (including time for memorization): 28 minutes 41.19 seconds - One of the earliest people in the world to achieve this.

Averaging around 2 to 3 minutes to blindfold solve a cube, timing starts when I first look at the cube and ends when I complete the solve. During the solving process, I do not look at the cube.