reSpeaker Introduction | Seeed Studio Wiki

[giscus[bot]]

reSpeaker Introduction | Seeed Studio Wiki

reSpeaker by Seeed Studio is an open-source modular voice interface platform for creating intelligent voice interaction products. Featuring advanced microphone arrays with far-field recognition, noise suppression, and audio processing algorithms, it delivers clear voice capture even in noisy environments. Compatible with multiple platforms and languages, it offers flexible development, open-source designs, and strong community support.

https://wiki.seeedstudio.com/respeaker/

[AzzA68]

A suggestive question, to the developers, from an old submariner and ex-sonar technician.

Do you plan on adding direction and range tracking, of sound sources, to your list of capabilities?

To do so ATM you could achieve X and Y plane tracking and ranging using phase analysis of each mic's output. You need only add code to do so. (This has been done using potato class hardware for decades, so I'd hazard a guess there's already enough CPU power using the Pi.)

Even better... to achieve full X,Y, & Z tracking, or full 3D, you would need to change the shape of the array. Either moving one of your current 4 mics into the Z plane, or adding a fifth mic in the Z plane. (Mics in a pyramid shape.)

Why?

e.g. This would enable a robot to turn in XY planes to face the sound source, and even tilt in Z to face its head/face directly at the sound source to eyeball it with its cameras. (This would be quite creepy, for most mere mortals, and only add more depth to the uncanny valley of robotics)

Have a look at Scout Subsea USBL Positioning, if you require some inspiration.

Great work already BTW!

[rayheto]

A suggestive question, to the developers, from an old submariner and ex-sonar technician.

Do you plan on adding direction and range tracking, of sound sources, to your list of capabilities?

To do so ATM you could achieve X and Y plane tracking and ranging using phase analysis of each mic's output. You need only add code to do so. (This has been done using potato class hardware for decades, so I'd hazard a guess there's already enough CPU power using the Pi.)

Even better... to achieve full X,Y, & Z tracking, or full 3D, you would need to change the shape of the array. Either moving one of your current 4 mics into the Z plane, or adding a fifth mic in the Z plane. (Mics in a pyramid shape.)

Why?

e.g. This would enable a robot to turn in XY planes to face the sound source, and even tilt in Z to face its head/face directly at the sound source to eyeball it with its cameras. (This would be quite creepy, for most mere mortals, and only add more depth to the uncanny valley of robotics)

Have a look at Scout Subsea USBL Positioning, if you require some inspiration.

Great work already BTW!

Thanks for the thoughtful suggestion and for sharing your experience! We’ve actually already planned follow-up products along this direction, including more advanced sound-source localization and tracking capabilities. Your ideas around 3D positioning and array geometry are very much aligned with where we’re heading. Stay tuned! 🚀