Sensory on Target with World's Most Accurate Low Cost Speech IC

7/7/2003 - Sensory, Inc., the world leader in embedded speech technologies, has announced that it is now taking orders and shipping its low cost, high quality RSC-4x family of integrated circuits. Designed to be the highest accuracy single chip speech recognizer, the RSC-4x also provides a powerful single chip solution for general purpose embedded controller applications, plus features such as speech and music synthesis, and biometric security through speaker verification.

"After over 2 years of development, the RSC-4x builds on our award winning RSC line-up by offering higher accuracy and more features while still hitting aggressive system cost targets." said Todd Mozer, CEO of Sensory. "We have combined everything our customers have been requesting over the past 10 years, including improved noise robustness, faster recognition response time, low-data rate speech synthesis, more product control features, and much better speech development tools."

Many Capabilities & Technologies
As a speech processor, the RSC-4x offers outstanding recognition capabilities including robust performance in noisy environments and a large vocabulary of words and phrases. The RSC-4x also provides high-quality speech synthesis at varying data rates from 3K-64K bits per second. Also included on chip is an 8-voice MIDI-like music synthesizer that supports voice overdubbing with music. Speech technologies include speaker independent, speaker dependent, speaker verification, and word spotting. Core speech modeling approaches include Neural Networks, Hidden Markov Modeling, Dynamic Time Warping, and other proprietary statistical and non-linear techniques.

Easy To Implement Tools!
The RSC-4x is supported by a complete suite of easy-to-use tools for both programming and speech processing. Programming tools are available from Phyton (, and include assembler, C-compiler, in-circuit emulator, and debugging/simulation tools, all designed for fast, bug-free programming under a user-friendly integrated development environment.

Sensory's revolutionary Quick T2SI (Text to Speaker Independent) technology allows text entry to create "push button" recognition sets that can be downloaded into an RSC-4x chip. The Quick T2SI tool supports easy customization for different vocabulary pronunciation requirements or foreign languages. This process dramatically reduces the time and risk required to develop quality embedded voice recognition products by lowering speech recognition set development time from months to minutes. Sensory's Quick Synthesis tool allows speech to be recorded and instantly compressed to varying playback rates using Sensory's proprietary SX synthesis.

One Chip, One Supplier
The RSC-4x enables exciting hands-free applications at very low system costs, with little more than a microphone, speaker, and resonator required for a complete application. Unlike many competing solutions, all technology software is included in the chip price, avoiding potential compatibility problems and negotiations with multiple vendors. Customers buying the chip gain access to all of Sensory's embedded technologies, and all technologies, speech data, and application code have plenty of room in the on-chip 128K bytes of ROM.

Hardware Specifications
The RSC-4x features an 8 bit micro-controller with DSP assist engines, twin DMA's, 16 bit ADC with microphone pre-amp, 10 bit DAC with PWM speaker drivers, as well as 24 configurable I/O lines including 2 comparator inputs. A real-time clock, general purpose timer, a watch-dog timer, a zero cost on-chip RC oscillator are also included, along with and low battery detection. The ADC, DAC, PWM, digital filter are all customer accessible for added features and control. The RSC-4128 offers 128KBytes of on-chip, maskable ROM, with address space for up to 1MByte off-chip. 4Kbytes of RAM is on chip with sufficient memory to store speaker dependent or speaker verification templates. ROM-less versions and various memory sizes are offered throughout the RSC-4x family.

Special hardware has been added to allow a low power audio wakeup mode (~100 micro-amps) where the chip can be awaken into a speech activity (synthesis, recognition, etc.) through a whistle or hand clap. This enables the RSC-4x processor to be the FIRST EVER battery operated recognizer with hands free usage (i.e. no button press to wake up!). The RSC-4x is designed to be ultra low power/low current, with <5 uA current in sleep mode, 100 uA in audio wakeup or idle mode, and 5-30 mA in general operating mode.

Price and Availability
The RSC-4x family consists of the RSC-4000 and RSC-4128. The RSC-4000 is a ROM-less version with address and data buses, while the 4128 contains 128 Kbytes of mask-programmable ROM. Pricing for the RSC-4x line starts at less than $2 in 100k quantities, which includes all required speech technology code.

Development and programming tools can be downloaded or ordered at Samples of the RSC-4x, evaluation & prototyping boards, and Speech Development Tools can be ordered now from

About Sensory, Inc.
Sensory, Inc. is the world leader in embedded speech technologies. Sensory's speech technologies include speech recognition, speaker verification, speech synthesis, and animated speech, and are used in consumer electronics, cell phones, PDA's, internet appliances, interactive toys, automobiles, and other environmentally challenging applications where low cost and high quality is essential. Sensory offers a complete line of integrated circuit and embedded software solutions to customers such as Avon, Fisher-Price, Hasbro, JVC, Kenwood, Matsushita, Mattel, MGA, Mitsubishi, Radica, Sega, Sharper Image, Sony, Tektronix, Toshiba, Uniden, and VOS. Sensory was founded in 1994 and is based in Santa Clara, California. More information is available from Sensory's web site at

Previous Page | News by Category | News Search

If you found this page useful, bookmark and share it on: