Some are very apparent as peaks in Figure 1. Such artifacts cause a lot of turning points and can falsely pretend stronger brain activity. Then there are muscle artifacts with rapid oscillations, greater than 30 Hz, often combined with changes of high amplitude, of duration 1s up to 20s and more. Low frequency perturbations with sufficiently high amplitude remove turning points, as explained for δ waves in Figure 2, and thus falsely indicate smaller activity. Our method is influenced by two types of artifacts. We choose d = 4.Īrtifacts are a common problem in EEG analysis. Figure 3: Continuous hypnograms of channel Fp2F4 of the subject of Figure 1 for different delays d = 1, 2, 4, 8, 16, corresponding to sampling frequencies 512, 256, 128, 64, and 32 Hz. For d = 1 we also had more than 6% of values with equal neighbors, which reduces to 1.7% for d = 4. On the other hand, d = 1 seems to be susceptible to small perturbations, since differences among neighboring values are so small. Turning rates for such d can exceed the level 2 3 of white noise. One should not take d > 10 for 512 Hz sampling frequency because then turning rates are strongly influenced by the presence of σ and α waves which cause the big variance for d = 16 in Figure 3. But the shape seems always the same, as long as we choose d between 1 and 8. The level of turning rates is smaller for small d, because of the smoothness of the series caused by the low pass prefilter.

Figure 3 shows the continuous hypnogram of Figure 1 for d = 1, 2, 4, 8, 16.

Hypnogram narcolepsy