My wife's pi head

Christines Man · Post by **Christines Man** » Mon May 23, 2022 11:40 pm

Technician wrote: ↑Sun May 22, 2022 4:33 am I changed gears a bit on my project. My family friendly one is good, but speech recognition is still less than perfect because I refuse to run one that sends data online for processing.
My doll is still using the Bluetooth speaker/mic. My new plan is to use an ESP module as a computer in the doll, theyre tiny, energy efficient, and no heat.
For the moment I'm stopped because I got a VERY time consuming new job, it's pretty cool though because I'm surrounded by robots. Still checking up on speech recognition periodically and I want to take another run at Vosk or DeepSpeech.

Congrats on the cool job. Robots!

I had some small success streaming audio from pi over to a python script running on a desktop, which played the audio into a virtual audio cable software, which then got picked up by Windows 10 speech recognition and voice attack which then would send words back to pi over http. Even though it kinda worked sometimes, I think the speech recognition drifted and it didn't work anymore. And omg, the vocabulary of Windows 10 speech recognition is too unnecessarily vast, and it will only allow me to exclude words. Really I'd like to exclude everything but I, love, and you. Piece of shit! So that was so convoluted and hacky. I have given up on speech recognition for a while.

Technician · Post by **Technician** » Tue May 24, 2022 6:10 am

Christines Man wrote: ↑Mon May 23, 2022 11:40 pm Congrats on the cool job. Robots!

I had some small success streaming audio from pi over to a python script running on a desktop, which played the audio into a virtual audio cable software, which then got picked up by Windows 10 speech recognition and voice attack which then would send words back to pi over http. Even though it kinda worked sometimes, I think the speech recognition drifted and it didn't work anymore. And omg, the vocabulary of Windows 10 speech recognition is too unnecessarily vast, and it will only allow me to exclude words. Really I'd like to exclude everything but I, love, and you. Piece of shit! So that was so convoluted and hacky. I have given up on speech recognition for a while.

Thanks, the new job is interesting but exhausting, 12hr shifts. I'm the only technician all night to keep the robots functional, easy some nights, naughty robots other nights.
I was thinking about the Google speech recognition that android phones have, it's very good and it runs without internet on an old phone I have here. So it makes me consider just using an SBC with Android.
Did you ever try those TTP223 touch switches? They're like 20 for under $9 on Amazon and I had good results with them through TPE. So cheap it doesn't even make sense to buy ttp223 chips and supporting components.

clicker · Post by **clicker** » Wed May 25, 2022 1:15 am

@ Christines Man

I have been experimenting with the touch sensor a bit. I took some of the insulated copper wire from a tiny motor and wrapped it around a Q-tip stick, covered it in epoxy, and that seemed to register my finger placed only about 1cm away.

How do you manage to insert them under the TPE without it making a bump? Do you have any pictures showing how you do it?

Technician · Post by **Technician** » Wed May 25, 2022 6:01 pm

When I did it I just put them between the mask and skull, they're so thin that I couldn't tell they were there. These are the ones I got, and I did experiment with cleaning a pad on the back where the integrated antenna is and soldering my own with mixed results: https://www.amazon.com/HiLetgo-TTP223-C ... NrPXRydWU=

Interplanetary · Post by **Interplanetary** » Wed May 25, 2022 7:01 pm

I am reading this - and I feel so many levels of inadequate. Holy heck - I admire your pasion and dogged determination to push this project forward and learn as you go. My best friend is an analog electronic enineer (and then some) I am half tempted to talk him in to joining this site JUST to view your posts and see if he might help you or offer tips or collaborate in some way. May I have permission to screenshot and send him your initial post to show him your work? (NO is a perfectly acceptable answer - I am just trying to help).

I am in awe.

Technician · Post by **Technician** » Sat May 28, 2022 4:19 pm

I wrote some ESP8266 code, and did some more TTP223 testing. I "think" (based on said tests) I can make some "way too sensitive" antennas and then dumb them down with 50pf trimmers to dial them in.
A C++ program I wrote polls the ESP, which is wifi direct (no router or internet involved) and will use tts through Bubbles bt speaker.
I'll be starting a new thread once hardware arrives and hopefully have a video soon.

Christines Man · Post by **Christines Man** » Mon May 30, 2022 7:54 pm

Interplanetary wrote: ↑Wed May 25, 2022 7:01 pm I am reading this - and I feel so many levels of inadequate. Holy heck - I admire your pasion and dogged determination to push this project forward and learn as you go. My best friend is an analog electronic enineer (and then some) I am half tempted to talk him in to joining this site JUST to view your posts and see if he might help you or offer tips or collaborate in some way. May I have permission to screenshot and send him your initial post to show him your work? (NO is a perfectly acceptable answer - I am just trying to help).

I am in awe.

Thanks! Yes. I don't mind at all.

Christines Man · Post by **Christines Man** » Mon May 30, 2022 8:10 pm

Technician wrote: ↑Wed May 25, 2022 6:01 pm When I did it I just put them between the mask and skull, they're so thin that I couldn't tell they were there. These are the ones I got, and I did experiment with cleaning a pad on the back where the integrated antenna is and soldering my own with mixed results: https://www.amazon.com/HiLetgo-TTP223-C ... NrPXRydWU=

These are very interesting. I may get some for testing. The disadvantage I see is the lack of control over sensitivity settings. Currently I'm able to capture raw capacitance and write my own thing to detect touches. An advantage is that you don't need to communicate with them over i2c so you could spread them out more and with really long wires.

So I have finished installing and coding some vagina sensing with voice response. Testing and will post some pics soon.

Technician · Post by **Technician** » Tue May 31, 2022 6:42 am

Christines Man wrote: ↑Mon May 30, 2022 8:10 pm The disadvantage I see is the lack of control over sensitivity settings.

Actually, there are two solder pads on the board. If a 50pf trimmer cap is connected there it becomes a sensitivity adjustment. I haven't tried it, but I plan on experimenting with that.
It's a SMD footprint, so I would like to put trimmers on another board, connected with shielded wire.

Could you explain a little about the MPR121 adjustable sensitivity? Like how adjustment works and wether you have an output that is on/off or variable?

Kaori Kusanagi · Post by **Kaori Kusanagi** » Tue May 31, 2022 7:35 pm

if you needed the "computer" to adjust the sensitivity, you could make a "digicap" like they show here
https://hackaday.com/2017/05/17/a-few-o ... nalog-mux/
(i've seen digipots but couldn't find a digicap)

Or a new one I just learned about
https://www.electronics-notes.com/artic ... -diode.php

Christines Man · Post by **Christines Man** » Thu Jun 02, 2022 7:19 pm

Technician wrote: ↑Tue May 31, 2022 6:42 am
Christines Man wrote: ↑Mon May 30, 2022 8:10 pm The disadvantage I see is the lack of control over sensitivity settings.
Actually, there are two solder pads on the board. If a 50pf trimmer cap is connected there it becomes a sensitivity adjustment. I haven't tried it, but I plan on experimenting with that.
It's a SMD footprint, so I would like to put trimmers on another board, connected with shielded wire.

Could you explain a little about the MPR121 adjustable sensitivity? Like how adjustment works and wether you have an output that is on/off or variable?

Ok, then that seems like it might be a better solution than MPR121. With the MPR121, I never end up using more than 3 or so sensor channels, and the rest goes to waste. It's got 12 channels.

I had read the datasheet for MPR121 very carefully, and found the chip doesn't seem to behave as the datasheet says it should. Specifically the way it handles adjusting the baselines. It was adjusting that baseline way too quickly, and adjusting the settings through I2C didn't seem to work at all. Unless I just gave up too soon, but I remember trying extra hard before going like, fuck this shit I'm out. It probably works great for something boring like a keypad. So I have resorted to fetching the raw values and doing the calculations in python code. It basically collects values when there's no activity, calculates the statistical mode, and sets that as the baseline. Works alright.

Christines Man · Post by **Christines Man** » Thu Jun 02, 2022 7:54 pm

So here's some images.

: Finished hair; 20220506_164125.jpg (208.4 KiB) Viewed 1427 times

Basically the result of my labors is looking like an expensive talking mop. I have been tying her hair back but she really needs a haircut at some point.

: Proximity sensor; 20220518_151032.jpg (90.85 KiB) Viewed 1427 times

This is an example of one of the touch sensors. This was the one that I made to see if 3 rows of wire is more sensitive than 1. It wasn't any better, so I am currently using one that is a single q-tip stick wrapped with wire a bit more tightly packed than this one, and coated in epoxy.

I may one day experiment with wrapping the wire, then wrapping that wire on top of the first wire. Or maybe try not round, more of a thin spade shape, but I didn't go there.

: Inserting wires 1; 20220518_191256.jpg (45.28 KiB) Viewed 1427 times

: Inserting wires 2; 20220518_191415.jpg (57.13 KiB) Viewed 1427 times

Ok, so I needed to route a USB cord from the pi down to where the sensor is, so I used a brass rod with something I could tie a thread to. This is kevlar thread. You can pull it as hard as needed and it won't break. Good for pulling stuff through TPE.

: Solder the sensor board; 20220518_201650.jpg (38.24 KiB) Viewed 1427 times

And then with the wire installed I soldered the wires. It's not a real USB port/cable. On the pi end I had installed a usb port where the leads are wired to the I2C pins.

: Attach sensors to board; 20220525_131029.jpg (106.39 KiB) Viewed 1427 times

I then pushed the sensors into place, positioned just above the vaginal cavity, as close as I could get it without poking through. There's a deep sensor, a midway sensor, and a clitoris sensor. The clit sensor is a nitinol "muscle wire" that is sticking out. I used nitinol because it's hair-thin, not sharp, and it will bend over and over and not break.

I used a cut up plastic bottle to hold the wound open, and started soldering. Added some epoxy to act as a strain relief, to keep the wires from moving around too much and breaking off.

I still need to glue up the wound but so far testing this, works great. She sounds pretty wild.

Technician · Post by **Technician** » Mon Jun 06, 2022 1:22 am

I experimented some more and made bus wire antennas. Coming off the TTP223 antenna, through a 47pf or 100pf cap, then a 100k variable resistor to ground allowed me to tune sensitivity.
I even had 2 antennas, to 2 sensors laying on top of one another (insulated between of course) and was able to have one sensor sensitive enough to catch a light touch ant the other desensitized enough to require a press. So i may do 2 sensors per region. They're only 50 microamps so what the hell.

Christines Man · Post by **Christines Man** » Thu Jun 09, 2022 2:05 pm

Technician wrote: ↑Mon Jun 06, 2022 1:22 am I experimented some more and made bus wire antennas. Coming off the TTP223 antenna, through a 47pf or 100pf cap, then a 100k variable resistor to ground allowed me to tune sensitivity.
I even had 2 antennas, to 2 sensors laying on top of one another (insulated between of course) and was able to have one sensor sensitive enough to catch a light touch ant the other desensitized enough to require a press. So i may do 2 sensors per region. They're only 50 microamps so what the hell.

I definitely think the TTP223 would work, but have you tried making a coil type of sensor to detect proximity? I was surprised how much that boosted sensitivity.

Sad to say, one of the coiled copper sensors has already broken off somewhere inside. I thought the strain relief would make it durable enough. From now on I'll not be using copper anywhere. I'm going to make a coil with muscle wire since that can bend back and forth thousands of times without breaking.

Technician · Post by **Technician** » Thu Jun 09, 2022 9:40 pm

Christines Man wrote: ↑Thu Jun 09, 2022 2:05 pm
Sad to say, one of the coiled copper sensors has already broken off somewhere inside. I thought the strain relief would make it durable enough. From now on I'll not be using copper anywhere. I'm going to make a coil with muscle wire since that can bend back and forth thousands of times without breaking.

Sensors see right through wire insulation, we could use silicone jacket wire that's very flexible, or wind copper inductors like you've been and coat them in hot glue for strength.
I'm trying to make area sensors that are 3" - 4" now to do her head first.