30 Jul '12
Oleg Kouzbit, Online News Managing Editor
Tatarstan-based RealSpeaker is trying to test the limits of computer vision based speech recognition—and hopefully push them. With a reported $148k already raised for what the developer believes is a disruptive technology the young Kazan IT Park resident is considering moving to Skolkovo outside Moscow and recruiting the capital’s best and brightest in computer vision. Under longer-term ambitious plans, the company wants to sell licenses for its technology in the US market, raise another $4m or more, and “refashion the way man interacts with the computer.”
RealSpeaker, a Kazan IT Park resident, has developed a technology that automatically recognizes a user’s speech of any duration and enables him to see it as plain text. The user, whether he speaks Russian or English, doesn’t have to adjust his manner of speaking to the machine, the developer says; the system is ‘taught’ to break speech down into separate fragments by analyzing lip movements and thus ‘reading’ information.
Skolkovo, Russia’s well-known innovation hub outside Moscow aimed to become the driver of national technology modernization, has recently backed the computer vision project by approving a $117k grant. A yet-unspecified co-investor is said to have pledged another $31k.
RealSpeaker says it will use the investment to hire more staff, arrange for alpha testing of its technology, and then work with a select group of up to 2,000 Internet users to test-run a Web-based version of the system.
If the innovative team pulls it off, the market may one day expect to embrace a technology that adjusts itself to a specific user instead of having him adapt to system functionality. RealSpeaker feels this would “refashion the way man interacts with the computer.”
Starting regional, then going global
The Tatarstan company is currently gearing up for its major next step. It is moving to Skolkovo in a bid to bring on board within the next six months what RealSpeaker founder and CEO Viktor Osetrov says will be “…the best team of computer vision experts in Russia.” To make this happen, the company is eyeing future space in a new Skolkovo-based office center slated for completion in September.
RealSpeaker is now headquartered in Tatarstan’s capital city of Kazan, some 500 miles east of Moscow. Its flagship speech recognition project began three years ago when the then 21-year-old founder, Mr. Osetrov, received his first grant. The core team includes five people, and some assignments are outsourced to other companies or individuals.
With new full-time staff the company will try and raise more funds to continue development. The project has been presented to Google, Intel and a number of US-based VC funds, and RealSpeaker says some are ready to invest as much as $4-4.5m after alpha testing is launched.
The Russian developer is interested in expanding to the US market and already set up earlier this summer in the United States a special company, RealSpeaker USA Inc., to look into new opportunities. The subsidiary will be instrumental in selling licenses for the Tatarstan technology all across America. The potential market includes audio speech recognition and voice navigation systems manufacturers.
RealSpeaker has also applied for a US patent.