TDD Mode Identification for Spectrum Sharing Applications

In recent years, spectrum sharing has emerged as a viable solution to satisfy the increased demand for spectrum resources owing to increased spectral usage. A cognitive spectrum sharing radar system implementing dynamic spectrum access (DSA) has been developed to address this need. To enhance DSA performance, we propose a method to optimize the time-division duplexing (TDD) mode of a 4G/LTE channel. This technique will cue the cognitive radar to anticipate times of low primary user transmission in allocated spectral bands, thereby allowing the radar channel to opportunistically access and utilize these bands. This TDD mode identification (ID) approach implements two simple machine learning algorithms. The data for classification are extracted from the LTE resource grid in the form of spectrograms. Fisher's linear discriminant algorithm and two iterations of K-nearest neighbor are used to perform the classification. The use of a majority vote step and optimization of the feature space size greatly increase the model efficacy. The model is successful while using only a small training dataset and is able to classify the TDD mode of a channel using only 10 ms of collected data. The overall classification rate achieved is 95%, with nearly 100% for frequency division duplexing band ID.