One of the most widely used methods for the analysis of biomedical signals is wavelet transformation. Covering a wide range of tasks, from compressing signals to finding correlations with mental functions, wavelet transformation algorithms have showed good results. Well-proven in electroencephalography (EEG) and electrocardiography (ECG) data processing, wavelet transformation nevertheless requires further enhancement focused on improving the detection of signal features relevant in an experiment or clinical study. This paper describes the method of creating filters that identify the features of biomedical signals based on the structure of chains of local extremums in the matrix of wavelet transformation coefficients. The paper also presents a method of selecting characteristic components related to operator activity from signals when working with Brain-Computer Interface (BCI) systems. There was also made a comparison with other methods of signal detection.