Performance Analysis of Logistic Regression Classifier for OTFS Symbol Detection

The application of machine learning techniques for wireless communication has grown explosively to support 5G and beyond. In this work, we use a logistic regression classifier (LRC) to address the orthogonal time frequency space (OTFS) symbol detection problem. For a static multipath channel, the LRC is trained in a way to detect the intended symbol. This method can be employed for parallel symbol detection. The performance of LRC under various channel scenarios is studied, and the results are reported in this paper. Simulation results show that for smaller size of OTFS channel, i.e., 4 × 4 and 16 × 16, the BER of the proposed technique is on par with the best of the discussed existing OTFS symbol detector. For a larger channel of size 256 × 256, the proposed technique detects with a promising accuracy of above 70%. The detection time of the LRC is lower compared with the existing techniques.