Page 91 - 2017 Brochure
P. 91
研究員
蘇 黎 Li Su
Assistant Research Fellow
Ph.D., Graduate Institute of Communication Engineering,
National Taiwan University
Tel: +886-2-2788-3799 ext. 1806 Fax: +886-2-2782-4814
Email: lisu@iis.sinica.edu.tw
http://www.iis.sinica.edu.tw/pages/lisu/
• Assistant Research Fellow, Institute of Information Science, Academia Sinica (2017-present)
• Postdoctoral Research Fellow, Center for Information Technology Innovation, Academia Sinica (2012-2016)
• Ph.D., Graduate Institute of Communication Engineering, National Taiwan University, Taiwan (2008-2012)
• B.S. (Double Degree), Electrical Engineering & Mathematics, National Taiwan University, Taiwan (2003-2008)
Research Description estimation (MPE) of polyphonic music.
I proposed a novel method, called the combined frequency and
My research interests focus on machine musicianship, the dream periodicity (CFP) method, which detects pitches according to
of making a machine understand music as a musician does. the agreement of a harmonic series in the frequency domain
Research areas for this interdisciplinary topic include automatic and a subharmonic series in the lag domain. This approach
music transcription, machine music appreciation, automatic music nicely aggregates the complementary advantages of the two
generation, and computational creativity of music. Applications are feature domains in different frequency ranges, and improves the
found in music production, education, entertainment, well-being, robustness of the pitch detection function to the interference of
computational musicology, revitalization of music cultural heritage, the overtones of simultaneous pitches. Experiments show that
and others related to next-generation music industry. Research the CFP method outperforms other algorithms for music data with
directions are two-fold: the first is the analysis of scores written by complicated instrumentation and a high degree of polyphony.
a composer (i.e., automatic music transcription), while the second The method also performs consistently well across all of the types
is the analysis of interpretation in musical performance (i.e., music of Western polyphonic music that have been considered, and is
expression analysis). robust to audio degradations such as high-pass filtering and MP3
compression.
As an ongoing project, the core technology I have developed over
the past three years is a system for automatic music transcription. symphony,” International Society for Music Information Retrieval
Automatic transcription of polyphonic music is arguably the Conference (ISMIR), August 2016.
holy-grail in the field of machine listening. Due to the difficulty of 7. Pei-Ching Li, Li Su, Yi-Hsuan Yang and Alvin W. Y. Su, “Analysis
creating pitch-labeled training data that covers the rich diversity of of expressive musical terms in violin using score-informed and
signals found in music, feature-based approaches derived from expression-based audio features,” International Society for Music
signal processing and feature design remain critical for multi-pitch Information Retrieval Conference (ISMIR), October 2015.
8. Yuan-Ping Chen, Li Su and Yi-Hsuan Yang, “Electric guitar playing
Publications technique detection in real-world recording based on F0 sequence
pattern recognition,” International Society for Music Information
1. Chen-Yun Lin, Li Su, and Hau-tieng Wu, “Wave-shape function Retrieval Conference (ISMIR), October 2015.
analysis – when cepstrum meets time-frequency analysis,” to appear in 9. Che-Yuan Liang, Li Su, Hsin-Ming Lin and Yi-Hsuan Yang,
Journal of Fourier Analysis and Applications. “Musical offset detection of pitched instruments: the case of violin,”
International Society for Music Information Retrieval Conference
2. Ping-Keng Jao, Li Su, Yi-Hsuan Yang and Brendt Wohlberg, (ISMIR), October 2015.
“Monaural Music Source Separation using Convolutional Sparse 10. Li Su, Li-Fan Yu and Yi-Hsuan Yang, “Sparse cepstral and phase
Coding,” IEEE/ACM Trans. Audio, Signal Language Proc. (TASLP), codes for guitar playing technique classification,” International Society
volume 24, number 11, pages 2158 - 2170, November 2016. for Music Information Retrieval Conference (ISMIR), October 2014.
3. Li Su and Yi-Hsuan Yang, “Combining Spectral and Temporal
Representations for Multipitch Estimation of Polyphonic Music,”
IEEE/ACM Trans. Audio, Signal Language Proc. (TASLP), volume
23, number 10, pages 1600-1612, October 2015.
4. Che-Yuan Liang, Li Su and Yi-Hsuan Yang, “Musical Onset Detection
Using Constrained Linear Reconstruction,” IEEE Sig. Proc. Lett.
(SPL), volume 22, number 11, pages 2142-2146, August 2015.
5. Li Su, Chin-Chia M. Yeh, Jen-Yu Liu, Ju-Chiang Wang, and Yi-Hsuan
Yang, “A systematic evaluation of the bag-of-frames representation
for music information retrieval,” IEEE Trans. Multimedia (TMM),
volume 16, number 5, pages 1188-1200, August 2014.
6. Li Su, Tsung-Ying Chuang and Yi-Hsuan Yang, “Exploiting
frequency, periodicity and harmonicity using advanced time-frequency
concentration techniques for multipitch estimation of choir and
89
蘇 黎 Li Su
Assistant Research Fellow
Ph.D., Graduate Institute of Communication Engineering,
National Taiwan University
Tel: +886-2-2788-3799 ext. 1806 Fax: +886-2-2782-4814
Email: lisu@iis.sinica.edu.tw
http://www.iis.sinica.edu.tw/pages/lisu/
• Assistant Research Fellow, Institute of Information Science, Academia Sinica (2017-present)
• Postdoctoral Research Fellow, Center for Information Technology Innovation, Academia Sinica (2012-2016)
• Ph.D., Graduate Institute of Communication Engineering, National Taiwan University, Taiwan (2008-2012)
• B.S. (Double Degree), Electrical Engineering & Mathematics, National Taiwan University, Taiwan (2003-2008)
Research Description estimation (MPE) of polyphonic music.
I proposed a novel method, called the combined frequency and
My research interests focus on machine musicianship, the dream periodicity (CFP) method, which detects pitches according to
of making a machine understand music as a musician does. the agreement of a harmonic series in the frequency domain
Research areas for this interdisciplinary topic include automatic and a subharmonic series in the lag domain. This approach
music transcription, machine music appreciation, automatic music nicely aggregates the complementary advantages of the two
generation, and computational creativity of music. Applications are feature domains in different frequency ranges, and improves the
found in music production, education, entertainment, well-being, robustness of the pitch detection function to the interference of
computational musicology, revitalization of music cultural heritage, the overtones of simultaneous pitches. Experiments show that
and others related to next-generation music industry. Research the CFP method outperforms other algorithms for music data with
directions are two-fold: the first is the analysis of scores written by complicated instrumentation and a high degree of polyphony.
a composer (i.e., automatic music transcription), while the second The method also performs consistently well across all of the types
is the analysis of interpretation in musical performance (i.e., music of Western polyphonic music that have been considered, and is
expression analysis). robust to audio degradations such as high-pass filtering and MP3
compression.
As an ongoing project, the core technology I have developed over
the past three years is a system for automatic music transcription. symphony,” International Society for Music Information Retrieval
Automatic transcription of polyphonic music is arguably the Conference (ISMIR), August 2016.
holy-grail in the field of machine listening. Due to the difficulty of 7. Pei-Ching Li, Li Su, Yi-Hsuan Yang and Alvin W. Y. Su, “Analysis
creating pitch-labeled training data that covers the rich diversity of of expressive musical terms in violin using score-informed and
signals found in music, feature-based approaches derived from expression-based audio features,” International Society for Music
signal processing and feature design remain critical for multi-pitch Information Retrieval Conference (ISMIR), October 2015.
8. Yuan-Ping Chen, Li Su and Yi-Hsuan Yang, “Electric guitar playing
Publications technique detection in real-world recording based on F0 sequence
pattern recognition,” International Society for Music Information
1. Chen-Yun Lin, Li Su, and Hau-tieng Wu, “Wave-shape function Retrieval Conference (ISMIR), October 2015.
analysis – when cepstrum meets time-frequency analysis,” to appear in 9. Che-Yuan Liang, Li Su, Hsin-Ming Lin and Yi-Hsuan Yang,
Journal of Fourier Analysis and Applications. “Musical offset detection of pitched instruments: the case of violin,”
International Society for Music Information Retrieval Conference
2. Ping-Keng Jao, Li Su, Yi-Hsuan Yang and Brendt Wohlberg, (ISMIR), October 2015.
“Monaural Music Source Separation using Convolutional Sparse 10. Li Su, Li-Fan Yu and Yi-Hsuan Yang, “Sparse cepstral and phase
Coding,” IEEE/ACM Trans. Audio, Signal Language Proc. (TASLP), codes for guitar playing technique classification,” International Society
volume 24, number 11, pages 2158 - 2170, November 2016. for Music Information Retrieval Conference (ISMIR), October 2014.
3. Li Su and Yi-Hsuan Yang, “Combining Spectral and Temporal
Representations for Multipitch Estimation of Polyphonic Music,”
IEEE/ACM Trans. Audio, Signal Language Proc. (TASLP), volume
23, number 10, pages 1600-1612, October 2015.
4. Che-Yuan Liang, Li Su and Yi-Hsuan Yang, “Musical Onset Detection
Using Constrained Linear Reconstruction,” IEEE Sig. Proc. Lett.
(SPL), volume 22, number 11, pages 2142-2146, August 2015.
5. Li Su, Chin-Chia M. Yeh, Jen-Yu Liu, Ju-Chiang Wang, and Yi-Hsuan
Yang, “A systematic evaluation of the bag-of-frames representation
for music information retrieval,” IEEE Trans. Multimedia (TMM),
volume 16, number 5, pages 1188-1200, August 2014.
6. Li Su, Tsung-Ying Chuang and Yi-Hsuan Yang, “Exploiting
frequency, periodicity and harmonicity using advanced time-frequency
concentration techniques for multipitch estimation of choir and
89
