High-frequency (HF) communications is undergoing resurgence despite advances in long-range satellite communication systems. Defense agencies are using the HF spectrum for backup communications as well as for spectrum surveillance applications. Spectrum management organizations are monitoring the HF spectrum to control and enforce licensing. These activities usually require systems capable of determining the location of a source of transmissions, separating valid signals from interference and noise, and recognizing signal modulation. Our ultimate aim is to develop robust modulation recognition algorithms for real HF signals, that is, signals propagating by multiple ionospheric modes. One aspect of modulation recognition is the extraction of signal identifying features. The most common features for modulation recognition are instantaneous phase, amplitude, and frequency. However, this paper focuses on two feature parameters: coherence and entropy. Signal entropy and the coherence function show potential for robust recognition of HF modulation types in the presence of HF noise and multi-path. Specifically, it is shown that the methods of calculation of coherence and entropy are important and that appropriate calculations ensure stability in the parameters. For the first time a new metric, called Coherence-Median Difference (CMD), is introduced that provides a measure of the dominance of coherence at specific frequencies to coherence at all other frequencies in a particular bandwidth.