Sunday, 5 April 2009

Voice recognition, on a future iPhone?

How much power would it need, if it were to delegate the learning to it's syncing computer?

Nuance owns:
T9 text input (Tegic - AOL owned for a while, but somehow went to Nuance)
Dragon Naturally Speaking (Wond a best of show at MWSF 2008)
And has it's foot in the medical secretarial door.


System requirements:
System Requirements
  • Intel-based Mac
  • Max OS X 10.5.1 or greater ("Leopard")
  • Internet connection required for product registration
  • 2GB of free hard drive space
  • A MacSpeech-certified USB noise-canceling microphone
http://www.nuance.com/speechplatform/

Audiomining http://www.nuance.com/naturallyspeaking/products/sdk/sdk_audiomining.asp

http://news.bbc.co.uk/1/hi/technology/7982763.stm

You'd need it open source - and ubiquitous . Seeing as you talk a lot on the phone, you'd be uniquely able to get a decent recording of key words in training, and modifier samples if you say something uniquely, through recording one side of the conversation.

Probably an ethical nightmare, but you could always upload a profile made on a desktop/laptop - or even do the 15 minute speech on the phone and then email to get analysed.