• Welcome to The Tolkien Forum! Connect with fellow Tolkien enthusiasts and join a community that is over two decades old, and still going strong. We look forward to meeting you.  Register

Doors of Durin jewelry box for my girlfriends engagement quest.

Callipygous

New Member
Joined
May 15, 2020
Messages
2
I made a Lord of the Rings themed jewelry box as part of a multi-stage geocache hunt for my girlfriend. It turns the door lights on when it's dark outside. It turns the mirror lights on when you open the doors.

Eventually I will have ring rolls above the drawers, mesh for hanging earrings in the empty spaces in the doors, and felted interiors on the drawers.

I am showing it to you in this unfinished state because I still have the hardest part left to tackle, and I hope a forum full of Tolkien fans might be willing to help. The latch in the left hand door is controlled by a raspberry pi concealed in the space behind the ring rolls. To really bring home the theme I'm going for, I am trying to get the RPI to open the latch when it hears the password. Since the pattern for the doors is Tolkien's original sketch for the Doors of Durin, the password is obviously the elvish word for friend: "Mellon".

I have been working with a voice recognition program designed for just such a purpose, but it has been totally unreliable. After I run through the training it works great. Picks up mellon, doesnt pick up melon, felon, fellow, mellow, etc. Perfect functionality... for about a minute. Then it stops responding completely. I kill the process and restart it. Now it picks up mellon, and hello, and melon, and felon, and when I knock on the table twice. Useless. I have been futsing with this program for months and can't get it to behave.

I have recently come across another option. A website, snowboy.kitt.ai, creates custom wakewords for similar applications. If they can get 2000 samples for a given word, they will produce a universal model that should be able to recognize the word as spoken by most voices. It asks you to sign in via social media, and then provide 3 spoken samples, test the personal model you've created, and then save. They seem to have some server issues. Right now it appears to be working, but only if I sign in via facebook. If you feel annoyed or put-upon by my request, please ignore and go about your day, I don't want to be a nuisance. But if you think its a cool idea and you would be willing to take a minute to help me make it a reality, I have provided a link below.

Thanks for reading!

Pictures:


Link to the hotword on snowboy:

https://snowboy.kitt.ai/hotword/16856

Link to pronunciation hint for "Mellon" Gandalf says it around 3:59

 
That is really cool! You know you’ll have to retrain the VR to account for accents. Perhaps raspberry pi is losing memory? Low battery, soft memory storage: it sounds as if it’s resetting. Is there anyway you can burn the memory, or at least prevent it from resetting?
 
Very nice! I have no idea what you smart magicians are talking about, but Raspberry pie sounds delicious.

Is there an Elvish word for doorknob ;)
 
Wow. This is just really cool, even without the voice command, which will make it even more awesome. I would lend my voice, but the link you attached isn't working.
 
Yeah, I kinda gave up on that approach. My friend found another piece of opensource software that lets you create a dictionary and quickly get some very basic speech recognition working. Lots of false positives, we hear it click open while were watching tv probably once a week, but it does open on command.

Thanks for the interest though!
 

Today's events

Back
Top