pyPhoneBooth

Can we use an old phone to capture speech and send it to a STT service? Can we use the same phone to receive text do TTS sound to the horn

Yes we can!

hardware

old phone

I used this one

microcoputer

RPi + audio codec

software

STT models

huggingface overview

RPi4 realtime STT

Whisper SOTA

RPi4 realtime STT

button off hook
  |
  v
play instruction message
  |
  v
record audio --> convert to correct samplerate
  |
  v 
send to STT service (over IP)

physicall connections

microphone input

When using electret microphone, I'm doubting if we need to have some gain on the electret side.

Audio comes in via MIC2_N, when using electret. MIC2_P when using onboard MEMS.

You need to use a coax cable coming from the electret to the stereo minijack (3.5mm) The tip is hot, the ring is cold, sleeve loose/cold.

button

When wiring the phone button to GPIO 27 (same as button on IQAudio), I get lot's of incorrect readings.

I've changed the phone button from GPIO 17 to GND, with internal pullup. This way, when the phone is on the hook, we get a positive reading on the GPIO (no shortcut to GND). When the phone is off the hook, we get a shortcut to GND and thus a 0 reading on the GPIO.

Even with this adaption, I get some bad readings.

TODO photo from internal connections of the phone & the IQAUDIO board TODO

I had some troubles creatingthe correct virtualenv

waveform80/rpi-lgpio#3

this fixed it:

I also ran into problems building by downloading from http://abyz.me.uk/lg/download.html

Turned out I had the same issue as gpiozero/gpiozero#1038 (comment), means my system-wide lgpio was Version: 0.2.2.0 and the virtual environment one was lgpio Version: 0.0.0.2.

I copied over following files from /usr/lib/python3/dist-packages/ to /home/pi/OctoPrint/lib/python3.11/site-packages/

lgpio-0.2.2.0.egg-info
lgpio.py
_lgpio.cpython-311-aarch64-linux-gnu.so

Now its working.

espeak or espeak -ng

makke sure epseak-ng is installed currently using subprocess/Popen to execute espeak

install

install as service

make a copy of the shell script phone.sh and name it phoneX.sh where X is the samne as the hostname index (ddd-phone1 > phone1.sh)
edit the new scriipt and adapt index
cp ddd-phone.service to /etc/systemd/sytem sudo cp ddd-phone.service /etc/systemd/system/ddd-phone.service
edit the service file sudo vi /etc/systemd/system/ddd-phone.service
enable the service sudo systemctl enable ddd-phone sudo systemctl daemon-reload

double check the sound card

(if it works from the login shell, this does not mean it will work from systemd) https://forums.raspberrypi.com/viewtopic.php?t=278665 https://forums.raspberrypi.com/viewto/questions/120034/python-script-not-playing-audio-when-run-through-systemd ddd-speaker.service requires /etc/asound.conf

The solution presented worked for me to get sound while running a python program from system.

What one has to do is to add defaults to the asound.conf file at /etc/asound.conf. I did sudo nano on file and found nothing in it, so I added

defaults.pcm.card 2

Why "2"? It's the number given when one does


pi@ddd-display1:~/ $ cat /proc/asound/cards

 0 [vc4hdmi0       ]: vc4-hdmi - vc4-hdmi-0
                      vc4-hdmi-0
 1 [vc4hdmi1       ]: vc4-hdmi - vc4-hdmi-1
                      vc4-hdmi-1
 2 [Headphones     ]: bcm2835_headpho - bcm2835 Headphones
                      bcm2835 Headphones

kaosbeat/pyPhoneBooth