TY - GEN
T1 - Sound source separation by spectral subtraction based on instantaneous estimation of noise spectrum
AU - Ozawa, Kenji
AU - Morise, Masanori
AU - Sakamoto, Shuichi
AU - Watanabe, Kanji
N1 - Funding Information:
This work was partly supported by JSPS KAKENHI Grant (JP19K04408).
Publisher Copyright:
© 2019 IEEE.
PY - 2019/11
Y1 - 2019/11
N2 - In our previous paper, we proposed a sound source separation method using the two-dimensional fast Fourier transform (2D FFT) of a spatio-temporal sound pressure distribution (STSPD) image that is composed from the outputs of a microphone array. In an STSPD image, vertical stripes are created for a target sound arriving from the perpendicular direction to the array; therefore, its spectral components are concentrated on the spatial direct current (DC) components in the 2D amplitude spectrum. In that study, we estimated the noise DC amplitudes using a deep neural network (DNN), then subtracted them from the observed spectrum to suppress the noise. However, the performance of noise suppression can be improved further. In this study, we estimate the noise DC components theoretically instead of empirically using a DNN. We improved the performance successfully.
AB - In our previous paper, we proposed a sound source separation method using the two-dimensional fast Fourier transform (2D FFT) of a spatio-temporal sound pressure distribution (STSPD) image that is composed from the outputs of a microphone array. In an STSPD image, vertical stripes are created for a target sound arriving from the perpendicular direction to the array; therefore, its spectral components are concentrated on the spatial direct current (DC) components in the 2D amplitude spectrum. In that study, we estimated the noise DC amplitudes using a deep neural network (DNN), then subtracted them from the observed spectrum to suppress the noise. However, the performance of noise suppression can be improved further. In this study, we estimate the noise DC components theoretically instead of empirically using a DNN. We improved the performance successfully.
KW - Instantaneous estimation
KW - Mathematical derivation
KW - Spatio-temporal sound pressure distribution image
KW - Spectral subtraction
KW - Two-dimensional FFT
UR - http://www.scopus.com/inward/record.url?scp=85081958093&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85081958093&partnerID=8YFLogxK
U2 - 10.1109/ICSAI48974.2019.9010477
DO - 10.1109/ICSAI48974.2019.9010477
M3 - Conference contribution
AN - SCOPUS:85081958093
T3 - 2019 6th International Conference on Systems and Informatics, ICSAI 2019
SP - 1137
EP - 1142
BT - 2019 6th International Conference on Systems and Informatics, ICSAI 2019
A2 - Wu, Wanqing
A2 - Wang, Lipo
A2 - Ji, Chunlei
A2 - Chen, Niansheng
A2 - Qiang, Sun
A2 - Song, Xiaoyong
A2 - Wang, Xin
PB - Institute of Electrical and Electronics Engineers Inc.
T2 - 6th International Conference on Systems and Informatics, ICSAI 2019
Y2 - 2 November 2019 through 4 November 2019
ER -