All News | Boards | Chips | Devices | Software | Archive | About | Contact | Subscribe
Follow LinuxGizmos:
Twitter Facebook Pinterest RSS feed
*   get email updates   *

Latest ReSpeaker mic array offers improved barge-in and full duplex operation

May 25, 2018 — by Eric Brown — 1809 views

Seeed has launched an upgraded, $69 “ReSpeaker Mic Array v2.0” for far-field voice control that works with any Linux, Mac, or Windows computer. It enables simultaneous playback and recording and offers improved barge-in voice recognition.

Seeed launched the ReSpeaker microphone array on Kickstarter in 2016, and followed up a year later with the $25 ReSpeaker 4-Mic Array for Raspberry Pi add-on board. Seeed’s new v.2.0 is an upgrade to the original ReSpeaker, but like the Raspberry Pi version, it’s not an SBC boardset, but is instead controlled via a separate attached computer.

ReSpeaker Mic Array v2.0 connected to laptop and speaker (left) and solo
(click images to enlarge)

You can connect the $69 device via its micro-USB port to any computer that supports USB Audio Class 1.0 (UAC 1.0), including all modern Linux, Mac, or Windows computers. Linux computers, such as the original, Linux-based ReSpeaker Core board, are recommended “for firmware updates,” says Seeed.

Like the original ReSpeaker the new version is a 70mm disc with 12x programmable LEDs. The ReSpeaker Mic Array v2.0 supports voice detection at up to five meters away “even with the presence of background noise,” says Seeed.

ReSpeaker Mic Array v2.0 and block diagram
(click images to enlarge)

Unlike the original, it’s a full duplex device in that that it enables a smart speaker device to “continue to play audio while also recording its environment.” It also provides barge-in support to quickly detect new voices or sounds when they enter the environment.

Thanks to improved MP34DT01-M digital microphones, the device integrates only four mics instead of the previous seven. The ReSpeaker v2.0 uses beamforming to separate out different voices, and it offers de-reverberation and noise suppression. The system provides -26 dBFS omnidirectional sensitivity, 63 dB SNR, and a 120 dBSPL acoustic overload point,

The original ReSpeaker combined a ReSpeaker Core board with a Mic Array board. The Core board had both an ATMega32U4 MCU and a 580MHz, MIPS-based Mediatek MT7688 processor running OpenWrt Linux. The mic board featured a XMOS XVSM-2000 microphone chipset.

ReSpeaker Mic Array v2.0 detail view
(click image to enlarge)

The v2.0 model is a single board that depends on an external computer for control operations, but processes audio using a more powerful XMOS XVF-3000 DSP chipset from XMOS. The XVF-3000, which also powers the XMOS xCore VocalFusion Speaker, provides improved algorithms including AEC (Acoustic Echo Cancellation), DoA, and BF.


Original ReSpeaker

The board is further equipped with a a 3.5mm audio output jack and a WM8960 stereo codec with Class D speaker drivers for 1W per channel audio. The board draws 5V power via the micro-USB port or an input header.

While the original ReSpeaker was billed as being hackable, Seeed mentions only that customers can customize microphone spacing to be configured as a circular or linear array. Seeed also says that customization services are available on request. Two firmware versions are available on GitHub, one including speech algorithms and a second for raw voice data.

Further information

The ReSpeaker Mic Array v2.0 is available for $69. More information may be found on Seeed’s ReSpeaker v2.0 shopping page and wiki.

(advertise here)

Print Friendly, PDF & Email

2 responses to “Latest ReSpeaker mic array offers improved barge-in and full duplex operation”

  1. William says:

    Also worth mentioning is the Acusis from Antimatter Research
    This uses the XMOS XVF3000 but is configured as a linear array.

  2. taqwa says:

    Can respeaker core v2.0 be used for full duplex acoustic echo cancellation: double talk detection?
    thanks in advance.

Please comment here...