基于阈值函数的诱发脑电小波特征增强

通过对脑-计算机接口中通信信号进行频谱分析,发现了靶刺激信号与非靶刺激信号的频谱在10Hz以下的低频段有较大的不同,采用Daubechies小波和Mallat算法对脑电信号进行多尺度分解,对高频分解系数简单置零、低频系数引入连续阈值函数进行滤波,使白噪信号在一定程度上得以滤除,靶刺激信号更加突出,提高了后续模式分类的正确率.

Feature Enhancement of Evoked Potentials via Wavelet Threshold Function

By spectrum analyzing to the brain-computer interface communication signals, it revealed that the spectrums between target and non-target signals are different significantly below 10Hz. The Electroencephalogram signals were decomposed in multi-scale by Mallat algorithm with Daubechies wavelet. The decomposition coefficients which related to high frequencies were simply set to zero, and a continuous threshold function was introduced to get the coefficients related to low frequencies filtered. The noises were filtered and thus the target signals were enhanced in some extent. The followed pattern classification accuracy was improved by these ways.