Identifying Neural Discharges using Time-Frequency Distributions for EEG

This paper presents a time-frequency approach as a nonlinear signal EEG processing technique. The proposed method is based on the use of the Smoothed Pseudo Wigner-Ville distribution (SPWV) good resolution combined with McAulay-Quatieri (MQ) sinusoidal model to identify a neural discharge. The initial results show the algorithm as a suitable method to develop an automatic detector based on graphics patterns parameterized by the features present in the neural discharges on the time-frequency plane. We obtained three features based on energy, frequency and tracking and the algorithm is tested in an application with epileptic EEGs. We can isolate a continuous energy trace with other oscillations when the epileptic seizure is beginning. This characteristic is always present in 16 different seizures from 6 epileptic patients.