Abstract
An objective assessment of the character cognitive process of human is presented, using statistical analysis of evoked potentials.
Alphabet patterns and random-dot patterns randomly chosen from among large varieties respectively are successively displayed in random order. Each pattern is displayed as 30ms flash at an interval of several seconds. Every EEG data is picked up for about 1 second, and 40 data of the same category are recorded. From each data optional 127 ms-length segment (128 samples) corresponding to a center latency is taken is put through Hanning window processing and fast Fourier transform to obtain the power spectrum whose fundamental frequency is 7.8Hz.
Then the two groups of 40 power spectrum each of the same center latency are put through discriminating analysis. Then examination is made to see in which frequency component and at what latency appreciable discrimination occurs. A method to evaluate the degree of contribution of each spectral frequency to the discrimination is also introduced, taking into account the fact that the power spectrum is correlated. The result of testing a healthy subject showed that there was an appreciable discrimination at 40-60Hz and latencies of around 130ms and 500ms. These are unavailable from conventional mean addition process. Thus, statistical analysis of wide frequency range VEP is one of the important directions for further study of evoked potential in general.
