Implementation of a neural network-based digital beamformer for a UMTS smart antenna

In typical UMTS mobile propagation environments, the interference level due to several user´s transmission is the main constraint factor of the system. Smart antennas are an efficient solution to reduce the interference level, thus increasing the number of users allowed in the system. The beamformer module of the adaptive antenna uses the information obtained from the array output to maximize the desired user´s signal and minimize the interference´s power, by properly combining the received signals. Typically, well-known adaptive algorithms are used to compute the optimum beamforming weights. An interesting alternative to traditional adaptive algorithms are artificial neural networks, specially for real-time implementation in a DSP. A digital beamformer based on a radial basis function neural network is presented in this paper, and its main features are studied. Its performance is compared to the results obtained with typical and optimal beamformers.