This is a "new", probably experimental, waveform associated with HEB and WLO. That particular example is probably from WLO, Mobile, Alabama. It is quite likely there were at least one, if not two, other frequencies active at that time with very similar, but not identical, waveforms. I am at work right now, so I do not have the exact frequencies handy, but one freq is in each of 14 MHz, 16 MHz, 17 MHz, and 21 MHz.
By the way, there is no "right" or "wrong" on this, but in general I call this signal on 17299 kHz. This places the MFSK signal within the normal audio passband of the transmitter used (there transmitter has been well characterized by other transmissions). Tuning to 17300 kHz will result in signal excursions below 0 Hz. Of course, if the highest and lowest frequencies of the signal were known you could designate by center frequency, and I think that might be about 17300.5 or so kHz.
T!
(edit) At home now, able to check the logs. I have logged this same signal on 16926.5, 17299, 17384, and 22819 kHz. Have received reports of it being on 14747 kHz but have not seen it there myself.
Something to note is the drift "trends" for these signals. While it might be on multiple frequencies at one time (I have seen up to 3 active at one time) they are not identical. However they do tend to go in the same direction. As one frequency drifts generally "up" in pitch so do the others, and when drifting down the same. Not step-for-step, but just kind of the same directions.