Advanced Speech Recognition System using Whisper AI

🚀 Abstract

This project showcases an Advanced Speech Recognition (ASR) System that transcribes audio to text in real time using OpenAI's Whisper AI model. With support for multilingual transcription, noise robustness, and a user-friendly web interface, this tool offers efficient voice-to-text applications for transcription, dictation, and voice-based user interfaces.

🔍 Motivation

Voice-driven technologies are revolutionizing industries like healthcare, customer service, and accessibility. Existing systems often face challenges in noisy environments, handling accents, or processing multiple languages. This project addresses these gaps by leveraging the Whisper model's versatility and accuracy to build a scalable, real-time ASR system.

🎯 Problem Definition

Traditional speech recognition systems are hindered by:

• 🚫 High latency
• ❌ Limited accuracy in noisy environments
• 🌐 Poor multilingual support
• 🔒 Privacy concerns due to reliance on cloud services

This project tackles these issues, aiming for real-time, accurate, and multilingual transcription, all while maintaining robust privacy measures.

🌟 Features

📝 Key Capabilities:

• Real-time Transcription: Accurate audio-to-text conversion.
• Multilingual Support: Dynamic recognition of multiple languages.
• Noise Robustness: Enhanced transcription in noisy environments.
• User-Friendly GUI: Intuitive web interface for seamless interaction.

🎯 Future Scope:

• Enhanced real-time conversion for time-sensitive applications.
• Support for additional use cases like subtitles and live translations.

🛠️ System Architecture

Workflow

1. Input and Preprocessing:

   • Context-aware transcription using past tokens.
   • Starts transcription pipeline upon receiving audio data.

2. Language and Speech Detection:

   • Automatic language identification for multilingual support.
   • Voice activity detection to filter non-speech audio.

3. Transcription and Translation:

   • Time-aligned Transcription: With timestamps for precise indexing.
   • Text-only Transcription: Continuous text output.
   • Optional translation from non-English to English.

4. Output Generation:

   • Generates time-stamped or continuous text.
   • Marks transcription completion per audio segment.

💻 Technology Stack

Software

• Streamlit: Web-based user interface.
• Whisper AI: Speech-to-text engine.

Deployment:

• Platform: Streamlit Cloud
• Deployment Link: Hertz ASR App

Networking and Infrastructure

• High-speed internet for real-time services.
• Scalable cloud-based infrastructure.

🚀 Getting Started

Clone the Repository

git clone https://github.com/AE-Hertz/speech-to-text/  
cd advanced-speech-recognition  

Install Dependencies

pip install -r requirements.txt  

Run Locally

streamlit run app.py  

🤝 Contributions

We welcome contributions to improve this project! Please:

1. Fork the repository.
2. Create a feature branch.
3. Submit a pull request.

📄 License

This project is licensed under the MIT License.

🛡️ Acknowledgments

• OpenAI for the Whisper model.
• Streamlit for the web development framework.

Happy coding! 😊