The spectrum access system project is part of the Defense Advanced Research Projects Agency (DARPA) challenge project at University of California, Berkeley. The aim of the DARPA spectrum challenge is to achieve a cooperative communication system. The Spectrum Access System team focuses on implementing an equalizer for the DARPA challenge team. The equalizer is introduced to reduce the distortion caused by intersymbol interference (ISI). The goals are to implement the equalizer in GNU Radio function block; assess the performance of combinations of different adaptive equalization models and adaptive algorithms.; Replacing the linear filter of Decision Feedback Equalizer with neural network and performance assessment on Decision Feedback Equalizer with neural network.
Software Implementation: The signed Least Mean Square, Normalized LMS, Variable-Step LMS and Recursive Least Square adaptive algorithms are implemented in Python to extend their applications into more telecommunication-related software (ex. GNU Radio).
Performance Assessment: The performance assessment is based on four metrics: steady-state error rate, convergence time, computational time complexity and computational space complexity.
Neural Network: The linear filter of equalizer is replaced by the neural networks. The structure of Neural network is composed of an input layer, a hidden layer and an output layer. The performance of common activation function is examined based on their convergence time.