TY - JOUR
T1 - Generalization of spectral flatness measure for non-Gaussian linear processes
AU - Dubnov, Shlomo
N1 - Funding Information:
Manuscript received October 30, 2002. This work was supported in part by the European Union Commission under CUIDADO project and Israel Science Foundation. The author is with the Ben Gurion University of the Negev, Beer-Sheval Israel (e-mail: [email protected]). Digital Object Identifier 10.1109/LSP.2004.831663 1The entropy rate can be considered as the average amount of bits per sample resulting from compression of an asymptotically long block of signal samples.
PY - 2004/8/1
Y1 - 2004/8/1
N2 - We present an information-theoretic measure for the amount of randomness or stochasticity that exists in a signal. This measure is formulated in terms of the rate of growth of multi-information for every new signal sample of the signal that is observed over time. In case of a Gaussian statistics it is shown that this measure is equivalent to the well-known Spectral Flatness Measure that is commonly used in Audio processing. For non-Gaussian linear processes a Generalized Spectral Flatness Measure is developed, which estimates the excessive structure that is present in the signal due to the non-Gaussianity of the innovation process. An estimator for this measure is developed using Negentropy approximation to the non-Gaussian signal and the innovation process statistics. Applications of this new measure are demonstrated for the problem of voiced/unvoiced determination, showing improved performance.
AB - We present an information-theoretic measure for the amount of randomness or stochasticity that exists in a signal. This measure is formulated in terms of the rate of growth of multi-information for every new signal sample of the signal that is observed over time. In case of a Gaussian statistics it is shown that this measure is equivalent to the well-known Spectral Flatness Measure that is commonly used in Audio processing. For non-Gaussian linear processes a Generalized Spectral Flatness Measure is developed, which estimates the excessive structure that is present in the signal due to the non-Gaussianity of the innovation process. An estimator for this measure is developed using Negentropy approximation to the non-Gaussian signal and the innovation process statistics. Applications of this new measure are demonstrated for the problem of voiced/unvoiced determination, showing improved performance.
UR - http://www.scopus.com/inward/record.url?scp=3543099304&partnerID=8YFLogxK
U2 - 10.1109/LSP.2004.831663
DO - 10.1109/LSP.2004.831663
M3 - Article
AN - SCOPUS:3543099304
SN - 1070-9908
VL - 11
SP - 698
EP - 701
JO - IEEE Signal Processing Letters
JF - IEEE Signal Processing Letters
IS - 8
ER -