LinTO-command-module

LinTO Command Module purpose is to:

• Spot wake-up-word
• Detect beginning and end of voice command utterances.

Introduction

linto-command module is a software brick for the LinTO Maker prototype. It is designed originaly to run on raspeberry pi 3 within LinTO-OS.

Its relies on several tiers libraries to acquire the audio from the microphone, detect voice activity, extract signal features and detect keyword using Tensorflow or Keras RNN models.

---

Getting Started

The easiest way to use linto-command-module is to download the binary package.

Binary package

The module has been tested on x86-64 and armv7l, binary archives can be downloaded there:

• x86_64: Here
• Armv7l (buster): Here

Extract the folder:

ARCH=armv7l
VERSION=0.2
tar xzf linto-command_$VERSION_$ARCH.tar.gz

From Source

1. Download the repository

git clone https://github.com/linto-ai/linto-command-module.git
cd linto-command-module

2. Install dependencies (w/ or w/out using virtualenv)

MY_VENV_FOLDER=path/to/your/venv
virtualenv -p /usr/bin/python3.X --no-site-package $MY_VENV_FOLDER
source $MY_VENV_FOLDER/bin/activate
pip install -r requirements.txt

---

Usage

• Executable from binary: EXTRACTED_FOLDER_LOCATION/command/command
• Executable from source: REPO/command/command.py

usage: command[.py] [-h] [--debug]

Voice Command Module

optional arguments:

  -h, --help  show this help message and exit

  --debug     Prompts debug

The executable comes alongside a .env_default file that contains its parameters. You can override those parameters using a .env file and you can overide those formers by using environment variables.

Exemple:

echo MODEL_FOLDER=/my/custom_model/path > .env && ./command --debug
                            or
MODEL_FOLDER=/my/custom_model/path ./command --debug
                            or
export MODEL_FOLDER=/my/custom_model/path && ./command --debug

Parameters are:

• MODEL_FOLDER: path to a folder containing a .net or .pb file alongside its .param file (default ~/model/kws)
• MQTT_LOCAL_HOST: MQTT local broker address (default localhost)
• MQTT_LOCAL_PORT: MQTT local broker port (default 1883)
• KWS_TH: Threshold for keyword detection (default 0.5)

---

Built using

• PyRTSTools - Custom tools for realtime speech processing.
• Tenacity - General-purpose retrying library
• paho-mqtt - MQTT client library.

License

This project is licensed under the GNU AFFERO License - see the LICENSE.md file for details.