All News | Boards | Chips | Devices | Software | Archive | About | Contact | Subscribe
Please whitelist in your ad blocker. Without ads from our sponsors, we cannot continue publishing this site. Thanks :-)

Speech recognition and synthesis shield runs Linux

Jul 6, 2015 — by Eric Brown — 1,741 views

Audeme has gone to Kickstarter to launch a “MOVI” Arduino shield designed for offline speech recognition and synthesis that runs Debian on an Allwinner A13.

A startup called Audeme has crossed the halfway mark on a $12,000 Kickstarter project for its MOVI (My Own Voice Interface) speech I/O shield for Arduino single board computers. An $80 early bird package ships in Feb. 2016, and a $100 package ships in Dec. 2015. The fundraising project ends on Aug. 10. The packages include access to low level serial interfaces that will enable developers to use MOVI with other SBCs such as the Raspberry Pi, says Audeme.

MOVI currently takes the form of a multi-board prototype, but the funding will enable a stackable single-board shield, says Audeme. The device won the Editor’s Choice Blue Ribbon Award at the Bay Area Maker Fair 2015.

Mockup of final MOVI shield (left) and current prototype
(click images to enlarge)

MOVI enables Arduino gizmos with voice-control for applications like turning devices on and off, entering alarm codes, and carrying on programmed conversations. The device offers voice synthesis and recognition for up to 200 customizable English sentences. MOVI is speaker independent, so there’s no voice training involved, and it uses no cloud services, thereby enabling offline applications, and without privacy or reliability concerns.

An onboard 2GB dictionary enables MOVI to “recognize virtually any complete English sentence as well as build their own dialog systems to change vocabulary,” says Audeme. The device is said to detect voices from up to 12 feet away.

MOVI runs Debian Linux on a Cortex-A8 Allwinner A13, which is the 1GHz Cortex-A8 system-on-chip that forms the foundation of the new Allwinner R8 SoC used in Next Thing’s $9 Chip SBC. MOVI is further equipped with an SD slot, which stores the Debian build, along with sentence, call sign, and configuration data.

MOVI’s programming interface
(click image to enlarge)

MOVI’s programming interface enables adding speech recognition to a project with fewer than 10 lines of code, says the company. MOVI also provides a programmable “call sign” for catching the device’s attention, and reducing false triggering.

An onboard electret microphone is available, but external headset mics are also supported. A 32-ohm audio output jack enables speech synthesis and optional feedback beeps, and a speaker is said to be optional. The device communicates with the Arduino board via digital pin 10 and 11, but you can use other pins using jumper cables and software adjustments.

MOVI prototype on the test bench
(click image to enlarge)

MOVI draws less than 3 Watts, which it obtains from the Arduino RAW power pin. The board does not require an external power supply, and can be powered by a battery. However, the Arduino board should probably be plugged into a battery or other power supply, “as most USB ports are not able to supply enough current,” says Audeme.

Further information

MOVI is available on Kickstarter through Aug. 10 for $80, for Feb. 2016 shipment, or for $100, for Dec. 2015 shipment. More information may be found on the MOVI Kickstarter page and at Audeme’s website.

(advertise here)


One response to “Speech recognition and synthesis shield runs Linux”

  1. Onekind says:

    I wonder which backend for the speech recognition is used.
    My guess would be pocketsphinx

Please comment here...