Polarization Dispersion Signal Feature Characterization for Preamble-Free Communications and Emitter-Based Classification

The polarization of a signal incident on a dual-polarized receiver is a function of the emitted polarization and the channel effects induced by the environment. Therefore, in a stationary multipath environment, the polarization of a signal from a transmit antenna to a dual-polarized receiver is constant, but can vary with frequency. This property can be exploited to define a polarization-mode dispersion (PMD) response that in most circumstances is unique for each transmit antenna in a multipath environment. This property is leveraged for signal classification in two applications. In one application, pulses from a network of frequency-hopping sources are classified according to their common source. Since the polarization response from a source is continuous in the polarization-frequency domain, signals in adjacent frequency hopping bins can be related if their polarizations are similar. By grouping frequency hopped signals based on their proximity in frequency and polarization, PMD responses for any number of unknown sources may be formed from collected data sets without need for training, and the signals can be classified and the confusion matrix estimated based on these responses. Furthermore, adding spatially separated receiver antennas with dual-polarized elements allows for space-polarization characterizations, where each possible pairing of receiver elements contributes a polarization-like signal feature that enhances signal classification.

Classification based on polarization features can also directly be applied for demodulation in orthogonal frequency-division multiplexing (OFDM)-based polarization modulated communications, where data are encoded into the emitted polarization of each tone in an OFDM symbol. This approach to demodulation does not require estimation of the channel effects across the entire symbol. Instead, by incorporating a sparse number of pilots to estimate the channel effects over a narrowband region of the symbol, demodulation can be achieved by identifying (i.e., classifying) the received polarization-subcarrier responses corresponding to each transmit polarization state. The use of pilots enables preamble-free demodulation and also enables a new method for carrier-frequency offset estimation which improves on other existing approaches reported in literature. Results are extended to hybrid modulations where information is encoded in the absolute phase of a polarization-shift keying tone with no loss of performance to the information encoded in polarization.