TPC investment to aid speech recognition company speed research and development

A new investment by Technology Partnerships Canada (TPC) in a small B.C. company could help solve the significant problem of noise interference in applications that rely on voice recognition. Wavemakers Inc. recently announced $4.4 million in new loans from the Industry Canada funding body, money that it will use to hire more research and development staff. The increased payroll will bolster Wavemakers’ chances of developing voice recognition software that is more accurate and requires less processor power within three years. The investment from TPC, which normally requires matching funds and in-kind contributions from the receiving company of around two-thirds of the cost of the total project, will be paid out of the agency’s Enabling Technologies envelope. As with other investments of this type, the money will be paid after claims made on allowable expenses, mostly associated with payroll. A media release from TPC indicates that 58 jobs will be created or maintained in the R&D phase of the project, and another 133 in the commercialization phase. Richard Sones, Wavemakers’ director of sales and marketing, cautions that the numbers published by TPC are optimistic. Still, he says, the work that will be undertaken is labour intensive, and the funding will help ease the human resources burden. Wavemakers is one of a handful of companies in the niche market of enhancing voice to improve the performance of speech recognition engines. The company currently sells a suite of products that can achieve a certain improvement in voice quality and block out interfering noise, but Sones says the industry is still facing the enormous challenge of enabling software to sort one voice out of many for the purpose of machine listening. Dubbing it the "cocktail party" problem, Sones explains that software can recognize the unique characteristics of human speech, but that it can’t yet recognize individual speakers – a serious problem for applications which use voice recognition interfaces in crowded areas. As an example, he cites voice-driven cell phone interfaces used in vehicles and elsewhere. When a car is crowded with chatting passengers, the dialing software can’t filter out legitimate commands from the driver from background noise. Sones recognizes the enormity of the challenge, but says the company has broken down the problem into manageable pieces, and has submitted a three-year schedule of smaller deliverables en route to reaching its broader goal. Still, he notes that the problem is a scientific one and the company will have to prove the basic science before it can create a commercial, practical product. "We’re optimistic. It is science – there’s no question about it. One of the things about science is that you don’t really know… You can have some good theories on how to approach it and how to solve it. Then you get into it, and you may come up with a solution, but it may, for example, not be practical. It may only work in a certain scenario because it doesn’t generalize, or it may not work in another scenario, or the code is way too big so it can never be used in the real world… "We certainly have some very good ideas within the group on ways of approaching it. We’ve been thinking about it for a while as a scientific group within the company. We’re going to attack it and if all goes well, we’ll nail it in three years." Currently, the company employs 21 people, of whom six are research scientists. It was established in 1996. Sones says the company’s most active clients are in the automotive market. While in-car telematics are a natural launching point for voice recognition interfaces, there are a number of other potential applications. These include potential uses in desktop computers and handheld devices, voice portals, and in other consumer electronics.

TPC investment to aid speech recognition company speed research and development

News | 01/23/2002 5:00 am EST

Unlock all the Canadian telecom, broadcasting and digital media news you need.

FREE TRIAL

SUBSCRIBE

Reuse & Permissions

Inside The Wire Report

Information

Hill Times Publishing

Get in Touch With Us