论文部分内容阅读
脑电图(EEG)分析已被广泛应用于疾病的诊断,针对癫痫患者的脑电检测可及时对患者的发病情况作出判断,具有很强的实用价值,因此急需癫痫脑电自动检测、诊断分类技术。为实现患者正常期、癫痫发作间期和发作期各时段脑电的快速、高精度自动检测分类,本文提出一种基于样本熵(SampEn)与小波包能量特征提取结合纠错编码(ECOC)Real AdaBoost算法的脑电自动分类识别方法。将输入信号的样本熵值和4层小波包分解后的部分频段能量作为特征,并用纠错编码和Real AdaBoost算法相结合的方式对其进行分类。本文采用德国波恩大学癫痫数据库实验数据(含正常人清醒、睁眼与清醒、闭眼,癫痫患者间歇期致痫灶外与致痫灶内及癫痫发作期5组脑电信号)进行了方法有效性检验。研究结果表明,该方法有较强的脑电特征分类识别能力,尤其对癫痫间歇期脑电信号识别率提升显著,上述5组3个时期不同特征脑电信号的平均识别率可达96.78%,优于文献已报道的多种算法且有较好稳定性与运算速度及实时应用潜力,可在临床上对癫痫疾病的预报及检测起到良好的辅助决策作用。
Electroencephalography (EEG) analysis has been widely used in the diagnosis of disease. EEG detection for patients with epilepsy can be timely judgment of the patient’s condition, has a strong practical value, so the urgent need of epilepsy EEG automatic detection, diagnostic classification technology. In order to realize the rapid and high-accuracy automatic detection and classification of EEG in the normal, seizure and episode phases of patients, this paper proposes a novel algorithm based on sample entropy (SampEn) combined with wavelet packet energy feature extraction and error correction coding (ECOC) Real AdaBoost Algorithm for EEG Automatic Classification and Recognition. The sample entropy of the input signal and the energy of the partial band after the 4-layer wavelet packet decomposition are taken as features, and the error-correcting codes are combined with the Real AdaBoost algorithm to classify them. In this paper, the University of Bonn, Germany epilepsy database experimental data (including normal sober, eyes open and awake, eyes closed, epileptic seizures in patients with interictal epileptogenic and intralesional epileptic seizures and epileptic seizures during 5 groups of EEG signals) method was effective Sexual test. The results show that this method has a strong ability to recognize and classify EEG signals, especially in the interval of epilepsy. The average recognition rate of EEG signals with different characteristics in three periods is 96.78% Which is superior to many algorithms reported in the literature and has good stability and speed of operation and real-time application potential, and can play a good role in decision-making and detection of epilepsy in clinical practice.