A control method of the prosthetic hand based on the electroencephalogram(EEG) of several designated mental tasks was proposed.The power spectrum density was estimated using Burg method to establish the multichannel EEG feature vectors.The vectors were fed into a linear neuron network to be classified.After the offline and online tests,a good result on recognizing 3 different modes was got.Using the enhancement of the α rhythm power when the eyes close as the triggering of the single test,the 3 degree of freedom control over the prosthetic hand was realized.