A voice controlled lamp using the continuous speech recognition engine julius and a ESP8266.

• personal computer (your average workstation should be sufficient)
• microphone
• ESP8266 (preferably a developement board like the NodeMCU)
• 5V relay module (those having externel circuitry onboard so they can be driven by a 3.3V signal)
• lamp of your choice, I used 12V high power LEDs, please do yourself a favor and avoid mains!

First you need to clone this directory and install julius.

git clone https://github.com/julianschuler/Voice-controlled-lamp.git
cd Voice-controlled-lamp
git submodule update --init
sudo apt install build-essential zlib1g-dev libsdl2-dev libasound2-dev
cd julius
./configure --enable-words-int
make -j4

Now you need to download the english DNN language model and unzip it.

cd ..
wget https://sourceforge.net/projects/juliusmodels/files/ENVR-v5.4.Dnn.Bin.zip
unzip ENVR-v5.4.Dnn.Bin.zip

Optionally, you can now remove the zip archive, since it is not needed anymore.

rm ENVR-v5.4.Dnn.Bin.zip

Make the scripts executeable.

chmod +x mic-test.sh voice-control.py./mic-test.sh

Now, test if your microphone and the voice detection works by talking to the microphone after starting the script.

./mic-test.sh

To use the python script, the package pyjulius is needed, it can be installed using pip (setuptools is needed by pyjulius):

pip install setuptools 
pip install pyjulius

Change STA_SSID and STA_PSK in the file lamp-code.ino to the WLAN credentials of your home network (important: The PC has to be connected to the same network) und save the modified file. Use the Arduino IDE and the ESP8266 core to upload it to the ESP8266. The relay module should have three pins on the input side, labeled "IN", "VCC" and "GND" or similar. Connect "IN" to GPIO14 (D5 for NodeMCU), "VCC" to 5V and "GND" to GND respectively. On the output side search for the normally open (NO) relay contacts, cut one of the power delivering wires of your lamp and hookup the contacts in between.

After everything is set up, execute the main script.

./voice-control.py

Wait a few seconds until the startup sequence is done, afterwards the lamp should turn on when saying lumos and off when saying nox. Note that the first input will be used for normalisation und thus always be discarded.

This project is licensed under the MIT license, see LICENSE.txt for further information.