Self-Organizing Map을 이용한 고잡음 신경신호의 클러스터링

Abstract

We present a result on unsupewised classification of extracellular neural signal with low signal-to-noise ratio (SNR) usrng selforganizing map {SUM). Instead of using simple amplitude thresholding, energy operator is used for the detection of neural spike to utilize both the instantaneous amplitude and frequency information. M e n applied to synthesized and real experimental data with high background noise, satisfactory clustering can be obtained. If necessary, learnxng time can be significantly reduced by using principal components of input data as the input to the map.