Epilepsy is a neurological disorder occurs at the central nervous system, Electroencephalography (EEG) is the reliable tool for analysing the human brain activity with the help of the signals, moreover, it plays a significant role in the detection of epileptic seizures. The abnormal electrical discharge leads to loss of memory; from the recent survey over five crore people are affected by epilepsy. An effective detection system is a vital solution for detecting the epileptic disease in the initial stage. In this paper, an improved epilepsy seizure detecting system is developed with a better accuracy; the EEG signal in both time and frequency domain with the use of Discrete Stationary wavelet-based Stockwell transform (DSWST) is proposed. The feature extraction is processed by a temporal feature, spectral feature and Amplitude Distribution Estimation (ADE) from EEG signals in which the normal EEG signals will have various spectral and temporal centroids. Also, a modified filter bank based particle swarm optimization (MF-PSO) helps for the feature selection; it significantly improves the classifier accuracy. Finally, a Hybrid K nearest support vector machine (Kn-SVM) is employed for classification to investigate the performance of feature to classify the brain signals into three groups of normal (healthy), seizure free (inter-ictal) and during a seizure (ictal) groups.