The RSC-4128 mixed-signal speechrecognition processor from Sensory has been available for a while. Yet designing a system around it has required a mix of analog and digital components and a significant amount of time. By preassembling all the components, from the microphone input to the control and speaker outputs, the VR Stamp module is a full subsystem in about the area of a large postage stamp (see the figure).
The RSC-4128 packs an 8-bit microcontroller, 16-bit analog-to-digital and digital-toanalog converters, digital filters, RAM, ROM, an output amplifier, timers, comparators, and more. The module's 1-Mbit flash memory holds the application software. Its 128-kbit serial EEPROM includes the speech library. Also, the module includes the main clock and real-time clock crystals, microphone input and speaker output, and power/noise management components. Designers can configure the module's 24 programmable I/O lines to respond to various voice commands.
Able to operate from 2.7 to 3.6 V, the module has an active current drain of 26 mA at 3-V and a sleep-mode current of less than 20 µA. As a result, designers can use it in a wide range of consumer and industrial applications, like set-top boxes, medical equipment, and elevators.
The VR Stamp simplifies the integration of speech recognition into products by combining all key components into a small, 40-pin dual-inline package (DIP) footprint. A full suite of development tools and the company's FluentChip Technologies speech modeling software make it easier than ever to integrate speech recognition and playback into a system.
A small programming and development card ties into a host system via two USB ports. The kit includes two VR Stamp modules and various software tools—FluentChip and Quick Synthesis 4, Quick T2SI Lite, and the Phyton MCC-SE integrated development environment. The tools run under the Windows NT, Windows 2000, and Windows XP operating systems.
The Quick T2SI (text to speaker independent) Lite is a special edition of the full T2SI tools. It permits speaker-independent vocabulary-set development using simple text input to create the desired commands. Each menu level of vocabulary can support up to about 40 words.
Designers can link an almost unlimited number of menus, depending on available memory. Combining the latest advances in neural networks and hidden-Markov modeling to create a powerful phonemic recognizer, the tool lets users create, edit, build, and download embedded vocabularies in minutes.
Quick Synthesis 4 compresses digital recordings of speech prompts in seconds. Additionally, it supports the easy scoring of MIDI-like music. International languages are supported, so end products can be deployed worldwide.
The complete VR Stamp toolkit costs $495, while the VR Stamp module sells for less than $30 each in volume. Both are immediately available.
Reprints
