Structural Complexity of Neural Signals by Matching Pursuits

Matching pursuits (MP) based adaptive approximation represents a signal with linear combination of a group of atom functions, which is closely related to the structural characteristics of the corresponding signal. In this study, we propose to use the average atom density (AAD) and the average atom scale (AAS) to evaluate the structural complexity of neural signals. For a simulated dynamic model, logistic map, both measurements behave similarly to the Lyapunov exponent spectrum. We apply them to the analysis of neural signal, namely EEG, from the brain with hypoxic-ischemic (HI) injury. The results show that AAD and AAS decrease in the early stage of the HI injury recovery because of the bursting and spiky activities. The preliminary results imply that the AAD and AAS can be used to describe the structural complexity changes in the neural signals, and can be used to segment the stages of HI injury and its recovery