francisricalde/thomas

new

Snowboy Hotword Detection: um pouco mais sobre a API

https://www.embarcados.com.br/hotword-detection/

Thomas: TJBot based assistant

Interact with tjbot in your own language, a voice based assistant fully customizable with a second screen. TJbot is an open source IBM's project. See this link for more information.

Hardware

To use this recipe you need 3 things:

• 1 Raspberry Pi 3 model B;
• 1 Speaker (P2 or USB entrance, it doesn't matter);
• 1 USB Microphone;
• 1 powerbank or any other kind of power supply;

link to the video >> https://youtu.be/jxWP9CqcVe0

This demo will need to be installed on your server and on your raspberry.

Part 0: Requirements

• Python3
• Python
• Node

To run this demo you need to have instances of the following services deployed on your Bluemix Account:

• Watson Assistant
• Watson Speech to Text
• Watson Text to Speech
• Watson Discovery (optional)

---

Part 1: Server

Clone the repository on your pc. Anywhere you like.

> git clone [email protected]:Leonardo.Henrique.Machado/thomas.git

If you don't have git on your pc, install it

The folder servidor100/ contains your server. Here you will recieve posts from your bot, and it will serve the screen that will be updated when it responds.

Go into servidor100/ and install the requirements:
Obs: Make sure pip is referring to your python3 installation

> cd servidor100/
> pip3 install -r requirements.txt

Maybe, on windows, the command will be only > pip install -r requirements.txt

To run the server, you should run the following command inside your servidor100/ directory:

python3 welcome.py

After that you can access your homepage by opening the url http://localhost:3000, or using the local IP Address http://<ip-address>:3000 so other computers can access it.

Displayed websites on second screen

You also need to update the intentions to point to the correct urls.

On line 10 of your servidor100/welcome.py file, you have a JSON object with a combination of keys and values, the key being an intent (defined by you on your own watson assistant service) and the value is a web site url.

Ex:

sites = {
    'wikipedia':'https://www.wikipedia.org/',
    'weather':'https://weather.com/'
}

You can use how many web sites you want in here.

Optional

If you want to enable Watson Discovery on this demo, you should edit line 60 of welcome.py and insert your credentials, environment ID and colletion ID.

---

Part 2: Raspberry

On the terminal, run:

> cd Thomas
> sudo chmod 744 setup.sh
> sudo ./setup.sh

Edit the following files with the credentials of the services you created before (Assistant, STT, TTS):
> Thomas/config.py

> Thomas/DynamicCaching/configureConversation.json

> Thomas/DynamicCaching/generateAudios.js
> Edit the lines "67, 68, 97 and 98"

Obs: If you need to generate new audios, you will need to delete a file called "sys_control" (cd /Thomas/DynamicCaching)

Then do this:

> cd Thomas/DynamicCaching
> python3 writeAll.py (run the file twice)
> node generateAudios.js
> cd Thomas
> python SnowBoyTest.py

Obs: You should add the ~/Thomas/inicial.sh to your crontab (@reboot ~/Thomas/inicial.sh) if you want it to start everytime with the operating system.

Now you have Thomas running, have fun!