We have brought forward a wavelet-based algorithm for electroencephalograph (EEG) signals--using scale dependent threshold based on median. In comparison with the universal threshold and Sure threshold, our proposed threshold, which is adaptive to the subband noise signals, preserves the noise free reconstruction property and takes lower risk than does the universal threshold; and our proposed threshold overcomes the drawback of Sure threshold. Evidently, the scale dependent threshold based on median is computationally simple and can obtain higher singal-to-noise ratio (SNR) it outperforms the universal threshold and Sure threshlold.
Algorithms ; Artifacts ; Electroencephalography ; Humans ; Signal Processing, Computer-Assisted