In this paper, we study the effect of tuning the
tunable-Q wavelet transform (TQWT) parameters on analyzing
the Electroencephalogram (EEG) signals used for detecting epileptic
seizure. Publicly available Bonn University database is used in
this study, fifteen different combinations were examined. TQWT
is used to decompose each EEG signal into a valuable set of band
limited signals (sub-bands), the value of the Q parameter is tuned
from one to four and the number of sub-bands (J) from six to
twenty two. Two statistical features were extracted from the subbands
having the highest percentage of total signal energy. Knearest
neighbor (K-NN) was used for classifying the EEG signals
into either seizure or seizure-free. Our results clarify that,
increasing the value of Q enhance the classification accuracy and
best results were achieved at Q equals two.