Automatic Speech Recognition

This repository contains the implementation of an Automatic Speech Recognition system in python, using a client-server architecture with Web Sockets.

If you want to know the explanation, I leave you the link to my video on YouTube. <a href="https://youtu.be/gdSUyI1z50o">YouTube: Speech Recognition in Your PC</a>

1. Files

• The docs directory the list of resources used for this project.
• The client.py script defines the client websocket. It handles stuff related to recognizing mic, setting audio features, etc.
• The server.py script defines the server websocket. It handles stuff related to loading Speech Recognition Models, inference, etc.

2. The architecture

<p align="center"> <img src='img/asr.jpg'> </p>

3. Dependencies

In order to install the correct versions of each dependency, it is highly suggested to work under a virtual environment. In this case, I'm using the pipenv environment. To install the dependencies you just need type:

pipenv install -r requirements.txt

then, in order to lauch the environment you would need to type:

pipenv shell

4. How to use

Once you have correctly installed the requirements. You must set in line 17 of client.py your input device. In my case, my device is defined as INPUT_DEVICE = "UMC204HD 192k".

Server

First, you need to launch the server. My recommendation is to use one terminail (or session) for the server. You can also run the server in background.

$ python -B server.py -l [EN | ES]

Client

Then, you will be able to lauch the client.

$ python -B client.py

5. Comments

Any comment, suggestion or colaboration, just reach me out at: fer.neutron@gmail.com

Feel free to clone or fork!