American Scientist

To the Editors:

I was interested to read Esther M. Morgan-Ellis’s section on mitigations to overcome internet latency for virtual choirs (“Virtual Community Singing During the COVID-19 Pandemic,” January–February). My research group at the NASA Ames Research Center has developed latency compensation for 10–20 milliseconds teleoperation lag using predictive filters. These techniques may well be applicable to audio signals to predict note, volume, and tone.

The predictive filter basically is a technique to use the history of a signal to guess where it will be in the near future. In the case of the voice signals, this kind of predictor would anticipate where the singers appear to be going musically in terms of note, timbre, and volume. It would then present listeners an advanced signal compensated for the time lag. In the case of virtual choir, the known musical score would be especially helpful because a predictor designer could use it to provide accurate prediction assistance.

Morgan-Ellis alludes to this kind of signal processing as being possible but requiring unusually fast computers. The clock on my six-year-old Apple laptop compares favorably with the machine that we used in our predictive filter experiments. Many even faster laptops are now available. Predictive filtering for virtual choirs should now be possible with widely available fast hardware, though it may need to be partially coded in assembly language for speed.

Stephen R. Ellis
Oakland, CA