Sensory is debuting FluentChip technology for the RSC line of microcontrollers. Also premiering is FluentSoft technology, which can be used on other hardware platforms. The products came to Sensory from the acquisition of Fluent Speech Technologies, a group of scientists out of the Oregon Graduate Institute's (OGI) Center for Speech and Language Understanding.
Dr. Pieter Vermeulen joined Sensory as part of the acquisition. The Fluent team, led by Dr. Vermeulen on the FluentSoft side and Dr. Robert Savoie on the FluentChip side, have spent over four years perfecting a technology that requires as little as 20K of memory to recognize a word set with extremely high accuracy, speaker independence, and reasonable tolerance for ambient noise.
"Sensory's reputation for excellence has been honed over the years, with customers with impeccable credentials for quality," notes Sensory CEO Todd Mozer. "Our new approach overcomes industry-wide obstacles by bringing down the cost of superior accuracy and noise tolerance to a point where high-quality speech recognition can be easily designed into even the lowest-cost lines of consumer products."
Sensory's RSC line of speech I/O microcontrollers now run the FluentChip firmware. These low-cost ICs sell for under $2 in volume, and can handle active vocabularies of up to 40 speaker-independent words. All the necessary ROM, RAM and hardware options are contained on chip, including a general-purpose microcontroller and powerful 16-bit A/D converter. By replacing an existing non-speech microcontroller, Sensory can add speech input and output, with an incremental cost of under $1. This opens up the technology to a new generation of consumer products, with user interface features that were never before affordable.
New language development tools, featuring FluentChip, cut speech development time from months to hours on the RSC platform. World class C-compilers, assemblers, debuggers and emulators offer inexpensive ways to prototype and implement speech recognition into any conceivable consumer application. FluentSoft is available as a complete Software Development Kit and is compatible with a number of popular DSPs, microcontrollers and various operating systems. FluentSoft is designed to work on a cell phone platform and can recognize over 600 names and continuous digits, even in a noisy environment.
Sensory develops embedded speech technologies, offering IC and software-only solutions for speech recognition, speech and music synthesis, speaker verification, and other voice and audio technologies.