Kohonen Winner Takes All Neural Network Realized on Microcontrollers with AVR and ARM Cores

This paper presents realization and the laboratory tests of the Kohonen winner takes all (WTA) neural network (NN) realized on microcontrollers (μC) with the AVR and ARM CortexM3 cores. Both μCs have been placed on a single testing board especially designed for this purpose. The board also contains an interface block with an analog-to-digital and digital– to-analog converters (ADC/DAC). The learning algorithm has been implemented on both μCs for the comparison. The board allows for switching between the μCs, between the Euclidean and the Manhattan distance measures. It also allows for turning on/off the so-called conscience mechanism. The learning process can be observed on-line due to DACs used on the board. Other network parameters can be viewed on PC using the USB port. The project is addressed to on-line analysis of the ECG signals in health care applications, as well as to the student laboratory.