The Lempel-Ziv (LZ) complexity and its variants are popular metrics for characterizing biological signals. Proper interpretation of such analyses, however, has not been thoroughly addressed. ⋯ We derive analytic expressions for the LZ complexity for regular and random sequences, and employ them to develop a normalization scheme. To gain further understanding, we compare the LZ complexity with the correlation entropy from chaos theory in the context of epileptic seizure detection from EEG data, and discuss advantages of the normalized LZ complexity over the correlation entropy.
Department of Electrical and Computer Engineering, University of Florida, Gainesville, FL 32611, USA. jhu@ecel.ufl.edu
IEEE Trans Biomed Eng. 2006 Dec 1;53(12 Pt 2):2606-9.
AbstractThe Lempel-Ziv (LZ) complexity and its variants are popular metrics for characterizing biological signals. Proper interpretation of such analyses, however, has not been thoroughly addressed. In this letter, we study the the effect of finite data size. We derive analytic expressions for the LZ complexity for regular and random sequences, and employ them to develop a normalization scheme. To gain further understanding, we compare the LZ complexity with the correlation entropy from chaos theory in the context of epileptic seizure detection from EEG data, and discuss advantages of the normalized LZ complexity over the correlation entropy.