reading about #pocketsphinx,
https://cmusphinx.github.io/2022/08/pocketsphinx-continuous/

author talks about #sox
>and it is so perfect that development on it largely ceased

BASED. THIS is the attitude that software sorely needs!
a project *completing* should be a good thing, but for some reason the neophilia is so strong that people can't imagine a state other than "incomplete" or "abandoned".

Ok..so my #ai #robotics project is just about done. I do these things as personal challenges and I've succeeded on all the major points, now just tweaking and putting them together.

I think I'll work on the speech response today and then integrate the movements and this will be a done one.

Here are some of the pieces parts for the nerds..
*Everything is local*

#Raspberrypi 4 2G
#Pca9685 i2c pwm controller
#Mg90s servos
#Pocketsphinx STT
#Festival light TTS
#python lots of python

@Corvus You can. There are several different ways to configure the wake word. You can use a predefined model, you can define a new wake word using phonemes and #PocketSphinx, or you can even train your own model using #Mycroft's Precise software. It's open source and provides the best accuracy. #Precise is based on a neural network that is trained on sound patterns rather than word patterns. It does take a lot more work though.

https://mycroft-ai.gitbook.io/docs/using-mycroft-ai/customizations/wake-word

Say Friend And Have This Box Open For You

Handcrafted gifts are special, and this one's no exception. [pender] made a Tolkien-inspired box for his son and shared the details with us on Hackaday.io. This one-of-a-kind handcrafted box fulfills one role and does it perfectly - just like with the Doors of Durin, you have to say 'friend' in Elvish, and the box shall unlock for you.

This box, carefully engraved and with attention paid to its surface finish, stands on its own as a gift. However, with the voice recognition function, it's a project complicated enough to cover quite a few fields at once - woodworking, electronics, and software. The electronics are laid out in CNC-machined channels, and LED strips illuminate the "Say Friend And Come In" inscriptions once the box is ready to listen. If you're wondering how the unlocking process works, the video embedded below shows it all.

Two solenoids keep the lid locked, and in its center is a Pi Zero, the brains of the operation. With small batteries and a power-hungry board, power management is a bit intricate. Two capacitive sensors and a small power management device are always powered up. When both of the sensors are touched, a power switch module from Pololu wakes the Pi up. It, in turn, takes its sweet time, as fully-fledged Linux boards do, and lights up the LED strip once it's listening.

[Pender] didn't want to go for any cloud-based voice recognition service - such a gift requiring a pre-established WiFi connection would be no fun at all. Instead, he set up and used PocketSphinx for offline voice recognition. The box is a surprise for his son, thus, [pender] couldn't quite ask for voice samples. Thankfully, PocketSphinx recognizes phrases without pre-training. There's also a contingency mechanism - the electronics aren't accessible until you open the box. The secret is to apply voltage to two pads, which will unlock the solenoids directly, just in case the battery dies or the Pi glitches out.

We don't quite know what awaits when the box is opened, but we can only hope that the son's name isn't Balin. We thank [pender] for documenting the details, as a reminder that our hacker skills let us build unique and beautiful things for the people we value. Hacker-built gifts are great - we've seen guestbooks for weddings, couple's proximity sensing LED badges, ESP32-based CTFs, and even meteorite turntable display stands.

#art #raspberrypi #hackadayio #pocketsphinx #powermanagement #voicerecognition #woodworking

Making Linux Offline Voice Recognition Easier

For just about any task you care to name, a Linux-based desktop computer can get the job done using applications that rival or exceed those found on other platforms. However, that doesn't mean it's always easy to get it working, and speech recognition is just one of those difficult setups.

A project called Voice2JSON is trying to simplify the use of voice workflows. While it doesn't provide the actual voice recognition, it does make it easier to get things going and then use speech in a natural way.

The software can integrate with several backends to do offline speech recognition including CMU’s pocketsphinx, Dan Povey’s Kaldi, Mozilla’s DeepSpeech 0.9, and Kyoto University’s Julius. However, the code is more than just a thin wrapper around these tools. The fast training process produces both a speech recognizer and an intent recognizer. So not only do you know there is a garage door, but you gain an understanding of the opening and closing of the garage door.

In addition, the tools are all made to work in Unix-style pipelines which is refreshing. Here's an example configuration from the project's website:

There are templating features so you can specify optional words and alternative words in a single rule. There are other features like mapping an object like living room lamp into something more computer-friendly.

Overall, this looks like a fun tool to have in your kit. If you do something interesting with it, be sure to drop us a tip so we can cover it. Meanwhile, we've been watching Linux speech for quite a while. Of course, what we really want is speech commands like the USS Enterprise, and we have to admit it is getting closer.

#linuxhacks #deepspeech #json #linux #pocketsphinx #speech #voice