Jump to content

Rotating steel ball and diaphragm


Andres Ramos

1,207 views

Today I finished a prototype where I incorporated all the findings and conclusions I gained from my previous experiments.

 

A steel ball is mounted to the rotating shaft of a gear motor. The motor is controlled by the SPICA circuit. The ball is pressed against the diaphragm on a tube. The tube is closed and a microphone picks up the signal from the inside of the tube.

The signal is post processed online with the lingua device.

I also implemented a hardware reverb module but it doesn't add much to the signal quality.

 

12 Comments


Recommended Comments

Intéressant votre système 🙂

Quand on enregistre le son que votre appareil fourni, en ralentissant un peu et bien il y a bel et bien des voix plus ou moins audible qui se font entendre. 🙂

 

Link to comment

Merci beaucoup Alain,


Ci-dessous, j'ai quelques images modifiées. Elles sont toutes en allemand.

Andrés

Link to comment

Excepter le quatrième son (dans le jardin) je n'entends pas la même chose que vous

- "Und dennoch machen wir Probleme" ->"Et pourtant, nous créons des problèmes"  (1- Ne t'étonne pas tout est bloqué)

- "Und Henning mag alle" -> "Et Henning aime tout le monde"  (2-  Henning, faut pas résister)

- "Und du machst jetzt Getriebe" -> "Et tu fais quelque chose avec une boîte de vitesse maintenant""  (3- On ne peut pas faire ce qu'il dit)

 

Link to comment

Merci beaucoup Alain pour tes interprétations. C'est merveilleux de voir que notre propre psychisme fait partie intégrante du processus de perception. Nous remarquons toujours que des personnes différentes entendent des choses différentes parce qu'elles parlent des langues différentes, mais que malgré cela, les messages ont souvent un sens.

Link to comment

Même en parlant la même langue, les gens entendront souvent autre chose que nous. Je pense que chacun entend ce qu'il doit entendre. Cela n'est qu'une supposition 😉

Link to comment

Today I tried Michael Lee's ML algorithm with my prototype and I was impressed. Generally the intelligibility is not satisfying yet but it sounds like real speech with tone characteristics and patterns. I recorded a 18m session were I continously optimized the ML parameters and the settings of my device. A lot of short fragments in German could be heard. Maybe you want to browse through it.

Recording from 2021-11-24

 

Link to comment

It definitely takes some tweaking. Also, I have "denoiser" ML model now that while still produces phonetic fragments, it sounds like the original voice that I trained it with. So everything will start to sound like a US General. 🤪

Link to comment
6 minutes ago, Michael Lee said:

It definitely takes some tweaking. Also, I have "denoiser" ML model now that while still produces phonetic fragments, it sounds like the original voice that I trained it with. So everything will start to sound like a US General. 🤪

Mr. Petreus, I guess. Yes pass me over that stuff.

Link to comment

I processed the results a bit more with increased pitch and decreased speed. When I set the noise gating threshold in the ML to the maximum that still produces results I get one voice like this.

With lower noise gating i get this type of signal with better voice characteristic but also sometimes with voices stacked on each other.

I tried Krisp for processing but it killed the whole signal, as usual. I think only Steve knows the Voodoo to tame Krisp. 😁

Link to comment
Guest
Add a comment...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

×
×
  • Create New...

Important Information

We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.