A straightforward method for calculating the voicing cut-off frequency for streaming HNM TTS

The Harmonic plus Noise Model vocoder produces natural text-to-speech synthesis without some of the artifacts encountered in other vocoders. However, in order to achieve this naturalness one needs to determine a voicing cut-off frequency for each frame of speech being synthesized. This has proven to be a challenge and there are many methods and implementations, all with certain trade-offs. We present here a straightforward method, based on cepstral energy, that can also be used in streaming HNM TTS synthesis.