Washington: Scientists are working on an advanced speech technology which they claim will be able to transcribe talks even in noisy conditions — useful for surveillance.
An international team, led by University of Western Australia, is developing the advanced system that will allow speech recognition in spaces filled with other voices, noise and echoes.
Team leader Prof Roberto Togneri said current speech recognition systems that turn a speaker’s words into text on a computer or mobile phone screen depend on the user wearing a head-set microphone or speaking close to the microphone in a relatively quiet environment.
So, the aim is to develop a system that frees the user from having to wear a headphone, or be right next to one, to provide on-screen text entry or voice control even in hostile, or noisy, multi-voiced, environments, the scientists say.
“We’re interested in developing advanced voice-activated control so that rather than a mouse or key pad, the user’s voice can activate controls such as ‘switch on’, ‘switch off’ and ‘turn the page’.
“We’re also interested in being able to position the microphone further away from the speaker. This will be useful in situations such as surveillance of criminal organisations and in information kiosks,” Prof Togneri said.