Regresar

Real-Time Display of Spiking Neural Activity of SIMD Hardware Using an HDMI Interface

Abstract:

Spiking neural networks (SNN) are considered the third generation of artificial networks and are powerful computational models inspired by the function and structure of biological neural networks, to solve different types of problems such as pattern recognition, classification, signal processing, among others. SNN have also aroused the interest of neuroscientists intending to obtain new knowledge about the functions of the neuronal system through the analysis of the patterns observed in spike trains. Therefore, in addition to the development of hardware solutions that allow the execution of the different neural models, it is important, to provide tools for the visualization and analysis of the spike trains and the evolution of the neural parameters of the affected neurons in real-time. This work describes a new solution that takes the hardware emulator of evolved neural spiking system (HEENS) as the starting point, which is a bio-inspired architecture that emulates SNN using reconfigurable hardware implemented in field-programmable gate arrays (FPGAs). Reported development includes new dedicated hardware modules to interface HEENS with the high definition multimedia interface (HDMI) port, ensuring execution cycles within a time window of at least 1 ms, a period considered real-time in many neural applications. Tests of the synthesized architecture including the new tool have been carried out, executing different types of applications. The result is a friendly and flexible tool that has successfully allowed the visualization of pulse trains and neural parameters and constitutes an alternative for the monitoring and supervision of the SNN in real-time.