Symbol Detection in presence of Symbol Timing Offset using Machine Learning Technique

Abstract

Orthogonal frequency division multiplexing is a multicarrier digital modulation technique that is extensively used in modern wireless communication systems. This technique is very sensitive to synchronization errors. Symbol timing offset is one of such synchronization errors. We here attempt to perform detection of symbols in presence of symbol timing offset using machine learning method. Symbol detection can be modeled as a classification problem. We use support vector machine method to classify the received symbols in one of many possible classes. We propose a special pilot data pattern that can be used to train multiple classifiers for different subcarriers and at different signal to noise ratios. We show that we incur lesser pilot overhead when we use this new machine learning based approach. A comparison between the traditional approach and our proposed technique has also been analyzed and presented.