Search

This paper addresses a novel machine learning approach to classify EEG signals that can be useful for discovering likelihood of suffering from genetic alcoholism. Excessive consumption of alcohol can create genetic cognitive impairment and contribute to alcohol disorder in descendants. Children can inherit it either directly from parents or any ancestors. Revealing chances of AUD would help in deciding preventive intervention efficiently. Major contribution of this work is classification based on input augmentation and automatic feature extraction. Instead of using signal from many spatial nodes proposed classification task relies on EEG signal from only two electrodes, therefore it is computationally efficient. Results are shown for performance comparison between customary and augmented EEG as inputs to the classifier. Input augmentation is confirmed for 15% improvement in accuracy.

This paper proposes a novel cylindrical double gate In0.53Ga0.47As vertical Nanowire n type device, which offers a higher drive current, better channel potential controllability and reduced short channel effects at ultra-short channel length 14 nm through numerical simulation approach. An independent inner gate control is used for threshold voltage tuning. Improvement of drive current in the Double-Gate Vertical Nanowire Transistor (DG-VNWT) structure is achieved by growing the In0.53Ga0.47As Nanowire vertically on silicon substrate. Further to reduce the leakage current of the structure, we propose high-k LaAlO3 outer gate stack in integration with Al2O3 passivation layer and WN metal gate for the first time. Variable threshold voltage tuning is also achieved by inner gate control of the proposed device. An analytical model is built for validating threshold voltage dependence on inner gate voltage. Our results of the analytical model are observed to be at par with TCAD simulations. The proposed DG-VNWT device is compatible with silicon and high-k/metal gate CMOS technology with the unique capability of inner and outer gate control, the most promising scalable candidate for low power applications.

Steganalysis of high capacity Wavelet based fusion image steganography with encryption, using Image quality metrics (as a set of features) is proposed. As the first order image statistics using the proposed algorithm are inherently preserved, which is desirable feature of the scheme, improving the security of algorithm against the targeted attacks .In addition comparing the present steganography scheme with two different encryption techniques, on the undetectibility ground, the generalized objective metric like SVD is used as a steganalysis tool. DFrFT encryption is found statistically and visually undetectable achieving the desired robustness though PSNR values are better in DNA encryption