Abstract
It is shown that contrary to the claims made in the above paper, the analytic signal frequency estimate can be severely biased. The bias manifests itself for short data records and low signal-to-noise ratios.