Power spectral scaling and wavelet entropy as measures in understanding neural complexity

The behavior of large ensemble of neurons in the brain is highly complex and very dynamic in nature, which necessitates the use of nonlinear methods for the analysis of EEG signals. Here we have addressed the issue of understanding the neural behavior in various brain states like eyes open, eyes closed, sleep states and the epileptic state by locating the variation of power distribution in the known frequency bands of EEG (such as beta, alpha, theta and delta activities) using the standard technique of Welch periodogram alongside approximate entropy which is a non-linear complexity measure of neural activity. We used both the raw signal as well as the wavelet decomposed signal using Daubechies wavelet at level five. The results indicate that approximate entropy is a more robust technique when combined with wavelet transform in understanding the complex method of brain process than that of spectral scaling methods derived from the slope of Welch periodogram.