Fujitsu Labs have developed a new system that allows modification of recorded voice parameters to change its apparent state of emotion.
Voice modulation can be very fun to do, with all that high comedic and low demonic pitch changing. However it does have some important uses, and for Fujitsu Labs, modifying recorded voice can be very serious business.
Normally, the only parameters that are modified in recorded voice are pitch and talking speed. While it can present a light or deep aura to the recorded voice, this type of modulation lacks the ability to convey importance, and is unable to stress urgency. To augment this incapability, Fujitsu Labs developed a new system that adds more modification parameters such as tone, intonation and (spoken) word spacing. A recorded voice modulated using this system could roughly convey a certain emotion or feeling, which can be adapted to suit the situation at hand.
For instance, the system could modify a fire emergency notification to make the voice sound more alarmed, increasing the urgency of the situation. The same system could also modify a traffic notification to make the recorded voice speak the words with a clearer tone and at a gentler pace. The system even includes an editing algorithm that would help modify any recorded voice automatically to the desired or needed emotional tone, by reading the words within the entire statement.
While the actual benefits of this system outside automated voice services are questionable, there is still a considerable practical potential to its technology. Its ability to infuse emotion can provide a significant addition to speech synthesis programs, or even to simpler computer-generated voices. Talking robots for example, could probably be made more lifelike with this system, and could possibly even take us deeper into the uncanny valley.
Source: Fujitsu Labs (JP)