Frame Synchronization In Ofdm Systems Using Neural Networks, With Dsp Implementation

Abstract

This paper presents a method for estimation of the timing offset in OFDM (Orthogonal Frequency Division Multiplexing) systems. The author showed that it is possible to detect the start of the OFDM blocks without the use of pilot tones or the most recently researched the use of correlation with cyclic prefixing, for frame synchronization. The method relies on training the neural networks to recognize a special synchronization pattern transmitted in the beginning of OFDM blocks. The tests and simulation carried out on the system showed an excellent recognition degree in an environment of high AWGN (Additive White Gaussian Noise) noise and multi-path distortion. Finally a real-time DSP (Digital Signal Processing) implementation of the method has been achieved with parallel results to simulation.