According to the feature waveforms of epileptic waves,a continuous processing system for epileptic waveforms detection is constructed based on the time-frequency domain localization of wavelet transforms.By detecting the local maximal positions across several successive dyadic scales of wavelet transform.the positions that indicate the points of sharp transitions in EEG are determined,thus the distribution of subwave periods in different scales can be obtained.Then by using the distribution,the epileptic waves in EEG are detected. This system shows good performance,such as high detection accuracy and capable of working continuously.Initial clinical results obtained were encouraging.