MCU Simulation Software Algorithm based on Deep Learning from Single Network Model to Multiple Layer Scenarios

Abstract

MCU simulation software algorithm based on deep learning from single network model to multiple layer scenarios is designed and implemented in the paper. For the realization of neural network in MCU, the selection and realization of the activation function is one of the key issues. Since the bipolar S-function has both a very good linear region and a very good nonlinear region, it can process both small and large signals at the same time. During the wiring process of the specific equipment, the audio and the video signals are 2-way signals, which must be connected to the input and output respectively, and the designed system is based on this operation. Therefore, in the software simulation system, it must also have these functions, and support the software compilation and also debugging environment. Furthermore, the designed model is also tested under the extreme scenarios.