Evaluation of the waveform in the magnetoencefalographic signals with noise in the form of a finite normal mixture

The article is devoted to the methods of evaluation of the waveform (or activation pattern) that occurs in the signals of the subject in response to an external stimulus during a neurophysiological experiment. The peculiarity of the mathematical model considered is the representation of noise in terms of finite normal mixtures. This approach is more consistent with the experimental data. Within the framework of the considered model, the analysis of different waveform estimators is carried out and their statistical properties are investigated. Several new estimators that utilize features of noise distribution were proposed. The author proved higher efficiency of new estimators in terms of variance amount in relation to the method of moments which is widely used in studies. Despite the fact that the article focuses on the application of estimators to the magnetoencephalographic signals, the considered methods can be used in similar applications for any neurophysiological signals as well as in the most general setting — for bias estimation of a finite normal mixture.