Epilepsy, a highly prevalent neurological disorder, profoundly impacts the lives of patients with periodic and unexpected seizures that lead to serious injury or death. This research work introduces a Siamese Convolutional Fire Hawk Sparse Autoencoder Network (SCFHSAN) to detect different states of the epileptic seizures. Initially, the EEG signals are gathered from two datasets including TUH EEG and UoB datasets. Then, artifacts removal is performed by multi-resolutional analysis and adaptive filtering technique. After that, dandelion tunable Q-wavelet transform is used to decompose signals into frequency sub-bands. Following that, the feature extraction and feature selection are processed using several techniques. Finally, Siamese convolutional sparse autoencoder network is proposed for epileptic seizure detection and fire hawk optimization algorithm is employed to optimize the weight parameter of the network. The results indicate that the introduced approach achieves an accuracy 99.95% and 98.78% on UoB and TUH EEG datasets, respectively. Analysis determines that the introduced SCFHSAN scheme would help neuro-experts in diagnosing epileptic behavior by analyzing seizure details from various brain regions using multichannel EEG signals.

癫痫是一种非常普遍的神经系统疾病，周期性和意外的癫痫发作严重影响患者的生活，导致严重伤害或死亡。这项研究工作引入了暹罗卷积火鹰稀疏自动编码器网络（SCFHSAN ）来检测癫痫发作的不同状态。最初，EEG 信号是从两个数据集收集的，包括 TUH EEG 和 UoB 数据集。然后，通过多分辨率分析和自适应滤波技术来去除伪影。之后，使用蒲公英可调谐Q小波变换将信号分解为频率子带。接下来，使用多种技术处理特征提取和特征选择。最后，提出了用于癫痫发作检测的连体卷积稀疏自编码器网络，并采用火鹰优化算法来优化网络的权重参数。结果表明，所引入的方法在 UoB 和 TUH EEG 数据集上的准确率分别为 99.95% 和 98.78%。分析确定，引入的 SCFHSAN 方案将通过使用多通道 EEG 信号分析来自不同大脑区域的癫痫细节来帮助神经专家诊断癫痫行为。