Current-mode implementation of discrete-time cellular neural networks using the pulse width modulation technique

In this paper, a novel implementation of discrete-time cellular neural networks (DT-CNNs) using the recently introduced current-mode pulse width modulation (CM-PWM) technique is presented. The CM-PWM technique can efficiently realize the delayed weighted summation required for DT-CNNs. All the weight currents are modulated to PWM pulses. Because of the translational invariance of a DT-CNN, a small number of weights (digital-like pulses) can be shared in the bus lines among all the cells. By changing the weight currents, a programmable DT-CNN can easily be achieved. This implementation of the DT-CNN results in small size of the whole system. This paper gives two application examples (a connected component detector and a hole filler).