Fujitsu Laboratories has today announced a new development in speech synthesis which will convey an appropriate vocal tone to messages spoken by your computer, workplace machinery, smartphone or tablet etc. The new tech provides speech synthesis in an "alarming tone," during an emergency, for example. Fujitsu says that its speech synthesis technology can communicate with users much more appropriately and effectively in a wide range of situations.
Fujitsu's work on developing speech synthesis software has cut the processing time to provide clear machine voice communications by approximately 1/30th of any previous technology. Due to this saving it is possible now to devote processing time to making speech sound more appropriate to the situation where it is used. This development doesn't just make speech synthesis more realistic sounding but can also help communicate more clearly in certain situations.
Multiple speech characteristics, able to be changed according to circumstances, go beyond the usual speed, pitch and vibrancy to include natural voice quality, intonation and pauses. Machine learning is used to apply the correct tonal characteristics to applicable situations. A number of applications for different kinds of voices are illustrated below.
Fujitsu foresees the synthesised voices, as above, changing if certain situations occur. For instance in the factory working example above, asking someone to shut off a valve; the first message would be in a clear relaxed tone of voice with volume and pauses depending upon the ambient noise level, then if an error occurred the speech could convey a "concerned tone", furthermore an emergency would be conveyed in a "very alarming tone". I think this could also work well in a car - with a sat nav combined with proximity and parking sensors.
Preserve your voice for eternity
This tech isn't all about alarming people though. "In addition, for all types of voice services, voices that perfectly match the customer's preference, such as voices that are perceived to be endearing, or distinctive voices for particular characters, can be used," according to Fujitsu. It is also foreseen that if a person was about to lose his or her voice, perhaps due to an illness, it could be recorded in advance and that person's speech synthesis would use their own voice as its foundation.