Comparative performance of time-frequency based EEG spike detection techniques

This paper investigates the performance of time-frequency based EEG spike detection techniques. The non-stationary nature of EEG makes time-frequency methodology a suitable tool for analysis. The high instantaneous energy of spikes are reflected as very localised energy patterns in the time-frequency domain with reduced time duration toward the high frequency area. These characteristics of spikes in the time-frequency domain makes them recognisable from the background. Two recently published spike detection methods, based on quadratic time-frequency and adaptive time-frequency parameterisation are considered in this investigation. These methods have been applied on both synthetic and real EEG signals. The obtained results show the superior performance of the quadratic time-frequency method for detecting EEG spikes in terms of resolution and robustness to noise.