Speech-to-Text using Gaussian Mixture Model & Hidden Markov Model

Description

Speech-to-Text is a technology that can convert voice data into text data. This allows computers to understand human language through voice commands. We combine machine learning technology into Speech-to-Text, namely the Gaussian Mixture Model and the Hidden Markov Model to identify sounds in text.

Live Demo Now is UNAVAILABLE.

Table of Contents

• Description
• General Info
• Technologies Used
• Features
• Screenshots
• Setup
• Usage
• Project Status
• Room for Improvement
• Acknowledgements
• Contact

General Information

Speech-to-Text is a technology that can convert voice data into text data. This allows computers to understand human language through voice commands. We combine machine learning technology into Speech-to-Text, namely the Gaussian Mixture Model and the Hidden Markov Model to identify sounds in text.

Technologies Used

• Flask==2.2.2
• hmmlearn==0.2.8
• ipython==8.10.0
• librosa==0.9.2
• numpy==1.23.5
• PySoundFile==0.9.0.post1
• python_speech_features==0.6
• scikit_learn==1.2.1
• scipy==1.10.0
• soundfile==0.11.0
• Werkzeug==2.2.2

Features

• Easy to Use
• Able to perform speech recognition and convert to text automatically

Lacks

• For now only supports English Language & File Format .wav

Screenshots

Setup

The requirements.txt file should list all Python libraries that needed for this project. This library will be installed using:

pip install -r requirements.txt

Usage

Type on your CMD or Terminal :

• Clone this Repository

git clone https://github.com/daffaarizkyy/STT_GMM-HMM

• cd to your directory (on where's you clone this project)

For Example:

cd STT_GMM-HMM

• Run pip install -r requirements.txt

• And Run python app.py

• Open your browser and enter localhost:5000 or http://127.0.0.1:5000/

Project Status

Project is: complete

Room for Improvement

Room for improvement:

• The Speech Recognition Processing needs to be improved so that the processing is more faster

To do for future development:

• Added more supported languages and file formats

Acknowledgements

• This project was inspired by Youtube Closed Captions and Many Films with Subtitle.

Many thanks to:

• Irvan Kurniawan : Modelling HMM Department of Informatics, Faculty of Computer Sciences, Universitas Sriwijaya, Indonesia

• Muhammad Daffa Rizky Fatarah : Modelling GMM and UI/UX Designer Department of Informatics, Faculty of Computer Sciences, Universitas Sriwijaya, Indonesia

• Osvari Arsalan, S.Kom., M.T : Lecturer and Researcher Department of Informatics, Faculty of Computer Sciences, Universitas Sriwijaya, Indonesia

• Rizki Kurniati, M.T. : Lecturer and Researcher Department of Informatics, Faculty of Computer Sciences, Universitas Sriwijaya, Indonesia

Contact

Created by @Wibu x Nolep - feel free to contact us!