Sensory Releases Speech Recognition Chip With OTP Memory
Sensory released the latest member of its speech recognition integrated circuit (IC) family, RSC-4192O, with one-time programmable (OTP) memory.
To compliment the new IC, Sensory also released an enhanced version of its firmware, FluentChip 3.0, to improve noise-immunity, speech and music synthesis, and multilanguage speech recognition accuracy.
Based on an 8-bit general-purpose microcontroller, the RSC-4192O offers 192 kilobytes of on-chip OTP memory. The RSC-4192O is a completely self-contained speech input/output (I/O) system with on-chip ADC, DAC, and output amplification.
The new FluentChip firmware includes enhanced speech features for better recognition accuracy in high-noise environments; new international lexicons and pronunciation predictors; and new acoustic models with 50 percent smaller code size, freeing much more of the RSC-4x processors’ code space for technology features in the end product. FluentChip 3.0 also includes Sensory’s T2SI (text-to-speaker-independent) technology, SD (speaker dependent; trained to one voice), T2SISD (allows both speaker independent and speaker dependent customized commands in the same command set), and SV (speaker verification; voice biometric password) speech recognition. A whole suite of technologies for enabling electro-mechanical animation is also included, such as Real-time LipSync, SoundSource tracking, and Beat Predict for dancing. FluentChip 3.0 also offers speech synthesis and MIDI-like music synthesis capabilities. Specialized tools are offered to facilitate fast prototyping and product development.
A new programmer tool from Phyton enables programming the RSC-4192O in 100LQFP package form at the customer’s lab.