Baidu may have found a way to dramatically advance speech recognition technology, according to a Forbes article by Patrick Moorhead. The Chinese web services company announced its progress via its Chief Scientist, Andrew Ng, at this year’s NVIDIA GTC 2015.
Essentially, Baidu has been using NVIDIA GPUs to expand a neural network supporting voice recognition processing, allowing the company to add massive amounts of data to the network so that its Deep Speech platform can use a more refined and effective deep learning approach to voice recognition. It’s an approach that could signal the start of a major overhaul in the technological infrastructure supporting voice recognition and biometrics technology, as the companies providing these services convert their data centers from the limits of current CPU processors to the advantages offered by GPU and FPGA chips. Baidu claims that with its new system, its Deep Speech platform has achieved 99 percent accuracy.
That could be a major advancement in the field that could greatly improve the speech recognition systems used in smartwatches and connected cars. And as the Internet of Things continues to expand and ever greater numbers of household devices acquire the ability to listen to their users, it could prove revolutionary in how we interact with the everyday objects in our daily lives.
—
March 30, 2015 – by Alex Perala
Follow Us