No abstract

Memristor resembles an artificial synapse and is considered to be basic electronic element for realizing neuromorphic circuits. In this work, a systematic investigation was conducted on memristor-based resistance-programming circuits to write analog data into a memristor utilizing pulse width modulation techniques. The high-frequency sinusoidal signal was utilized to read the data in the form of its electronic resistance. An optimum circuit configuration demonstrated multilevel stable resistive states, which are analogous to the connection weights in the human synapse. In order to modulate these memristive weights for representing the learning activities in human brain synapse, it was identified that the pulse width modulation technique is superior as compared to spike-timing-dependent plasticity. Further, the above analysis was utilized in training the memristor to update its resistive weights in consonance with its learning, analogous to that in a neural network. Further, the memristive crossbar architecture was utilized to implement a real-time application in Econometrics, where an array of memristors were utilized to learn and update the purchase trends of an [Formula: see text] matrix of customers. The proposed circuits possess the advantages of high packing density, low power consumption and nonvolatility, and also pave the way for developing future neuromorphic circuits.

Remote detection of the cardiac pulse has a number of applications in sports and medicine, and can be used to determine an individual's physiological state. Previous approaches to estimate Heart Rate (HR) from video require the subject to remain stationary and employ background information to eliminate illumination interferences. The present research proposes a spectral reflectance-based novel illumination rectification method to eliminate illumination variations in the video. Our method does not rely on the background of the video and is robust to extreme motion interferences (head movements). Furthermore, in order to tackle extreme motion artifacts, the present framework introduces a novel feature point recovery system which recovers the feature tracking points lost during extreme head movements of the subject. Finally, the individual HR estimates from multiple feature points are combined to produce an average HR. We evaluate the efficacy of our framework on the MAHNOB-HCI dataset, a publicly available dataset employed by previous methods. Our HR measurement framework outperformed previous methods and had a root mean square error (RMSE) of 5.21%.

Independent component analysis (ICA) is an unsupervised learning approach for computing the independent components (ICs) from the multivariate signals or data matrix. The ICs are evaluated based on the multiplication of the weight matrix with the multivariate data matrix. This study proposes a novel Pt/Cu:ZnO/Nb:STO memristor crossbar array for the implementation of both ACY ICA and Fast ICA for blind source separation. The data input was applied in the form of pulse width modulated voltages to the crossbar array and the weight of the implemented neural network is stored in the memristor. The output charges from the memristor columns are used to calculate the weight update, which is executed through the voltages kept higher than the memristor Set/Reset voltages (±1.30 V). In order to demonstrate its potential application, the proposed memristor crossbar arrays based fast ICA architecture is employed for image source separation problem. The experimental results demonstrate that the proposed approach is very effective to separate image sources, and also the contrast of the images are improved with an improvement factor in terms of percentage of structural similarity as 67.27% when compared with the software-based implementation of conventional ACY ICA and Fast ICA algorithms.

No abstract