Browsing by Author "Kadhim, Yezi Ali"
Now showing 1 - 1 of 1
Results Per Page
Sort Options
Item DEEP LEARNING APPLICATIONS IN SEIZURE DETECTION(2022-02-24) Kadhim, Yezi Ali; Mishra, Alok; Garg, LalitIn this thesis, a new method is proposed based on deep auto encoder and power spectral density. First, the input data is analyzed using power spectral density for feature extraction by measuring the power spectral density of the signal for each row of data. The produced output becomes input to the first Auto encoder to reduce the dimension and extracted high level features. The output of first auto-encoder become input to the second auto-encoder also to reduce number of features and extracted high level features. In addition, these features are classified into two groups: normal and abnormal by using SoftMax classifier. Finally, the two auto-encoders and SoftMax stacked and trained by using backpropagation algorithm to improve the classification accuracy. The proposed method gives satisfactory results when compared with the common methods presented in this filed .Here, the number of Auto encoders depend on the behavior of the data as well as the dimension. The proposed method is tested with commonly used datasets in the epilepsy serius detection, and the results obtained are compared with other and most prominent works in this field in order to determine the strengths and weaknesses.