Design of neuron architecture on FPGA for electromechanical sensor signals / Khairudin Mohamad

Artificial neural networks (ANN) are known to be able to improve electrochemical sensor signal interpretation. The hardware realization of ANN requires investigation of many design issues relating to signal interfacing and design of a single neuron. This report focuses on the design of neuron architecture on FPGA for electrochemical sensor signal. The objective of this project is to translate the data from electrochemical sensor signals and process the data with neuron structure and analyze how different digital module of the neuron could affect the data accuracy and performance of the design. It encompasses interfacing from analogue to digital, data structure and the design process of the simple neuron which includes adder, multiplier and multiplier accumulator (MAC). A major component of the algorithm is the design of the activation function. The chosen activation function is the hyperbolic tangent which is approximated by Taylor Series expansion. The neuron is evaluated on an Altera DE2-70 FPGA. The performances are evaluated in terms of functionality, usage of resources and timing analysis. For the data structure, it was demonstrated that increasing the fractional bits increases the precision. For the MAC was found that by using topology 2, propagation can be reducing up to 19.145ns