YUAN-PING CHEN

Music and Audio Computing Lab, Academia Sinica, Taipei, Taiwan
"Automatic Transcription of Lead Guitar in Real-World Recordings"

• Developed a web-based application to transcribe guitar solos in mixed recordings into sheet music

• Implemented real-time foreground source separation algorithm to isolate guitar solo from instrumental accompaniment by exploiting repeating structure of music

• Removed percussive source from musical mixture by time-frequency masking

• Developed a three-state left-to-right hidden Markov model (HMM) to recognize musical notes in pitch contour by modeling the temporal behavior of pitch values during a note event

• Improved guitar expression style recognition using musical data augmentation methods
   

Music and Audio Computing Lab, Academia Sinica, Taipei, Taiwan
"Expression Style Recognition of Lead Guitar in Real-world Recordings"

• Detected 5 electric guitar playing techniques in real-world guitar solo recordings using time series pattern recognition and machine learning techniques, achieving average F-score 0.74

• Implemented feature extraction, sparse coding, dictionary learning and classification method

• Separated guitar audio signal from mixture using robust principal component analysis (RPCA)

• Implemented melody extraction algorithm to estimate the melody contour from polyphonic music

• Established a new electric guitar solo dataset comprising 42 phrases and 1,134 musical notes

• Proposed new set of features to classify guitar playing techniques

Ultrafast Dynamics Lab, NCTU, Hsinchu, Taiwan
"Ultrafast Dynamics in Bi2Se3, CuxBi2Se3, MnyBi2Se3 Studied by Femtosecond Pump-Probe Spectroscopy"

• Designed optical path and pump-probe system to enhance time resolution of femto-second laser

• Investigated ultrafast dynamics in Bi2Se3, CuxBi2Se3, MnyBi2Se3