Established in 2004 by a team of audio signal processing researchers from the University of Southern California, Audyssey spent more than a decade studying various listening environments to improve the audio listening experience in traditional settings (home, cars, movie theaters, etc.). The purpose of that research was to achieve a keen understanding of sound reflections, what caused them, and how to correct them to ensure that audio is perfect when it reaches the listener’s ears. Now, Audyssey announced its eVR suite of audio technologies for far field voice recognition in CE products that rely on voice-based user interaction.
“Today, with a new generation of voice-activated CE products on the horizon, tiny microphones embedded in the devices themselves will be trained to hear and interpret voice commands from the user. If the “ears” in these new products aren’t hearing commands correctly, products won’t function correctly and users will become frustrated,” the company says.
Audyssey's eVR Suite allows for devices to better understand words from a significantly farther distance, up to 6 meters/19 feet. Word recognition rate remains high from various distances, even in a reverberant room. And with Audyssey's Barge-In technology, devices can detect spoken keywords, even when the device is playing at maximum volume. Manufacturers that license eVR can achieve accurate and robust far field voice recognition in their products, thus delivering a more enjoyable customer experience.
Approximately six billion products with microphones and voice interaction will reach consumers by the year 2020. Input voice quality will be a key enabler of this new user experience. With the first wave of voice-controlled CE products now in consumers’ hands, it is increasingly clear that, while these products respond well when the user is within close proximity of the device, the user experience deteriorates rapidly when interacting with the device from greater distances.
“Far field performance has been the Achilles’ heel of voice recognition, and has up to now been a barrier to the proliferation of voice-controlled hardware in the IoT ecosystem,” says Chris Kyriakakis, Audyssey’s founder and CTO. “Audyssey eVR is the first complete far field solution that gives CE manufacturers an easy way to leapfrog their competitors by delivering unparalleled voice recognition performance.”
Leveraging the company’s 10+ years of research in room acoustics, sound reflections and reverberation, Audyssey researchers invested two additional years of R&D characterizing the problems of far field voice capture and developing new algorithms to capture and clean up voice signals so they match what the voice recognition engines in the cloud are expecting to hear. Voice-controlled devices must also be alerted to start listening to user commands. Audyssey developed a unique method to create customizable “hot words” that can be used to wake up devices even when they are playing loud music. Extensive benchmark testing reveals that Audyssey eVR improved word recognition by more than 50% compared to popular front-end capture devices, and barge-in performance was improved by more than 70%.
Audyssey eVR is a lightweight software solution, easy for CE manufacturers to incorporate into their products and services. Hardware manufacturers interested in licensing Audyssey eVR can work directly with Audyssey’s engineers to optimize the microphone configuration based on the functional requirements of the product, and customize the hot word to meet the user experience requirements. This makes it possible for CE manufacturers to deliver the voice recognition performance their customers expect in any room or listening environment.
www.audyssey.com
