Stuttering Detection and Correction Using Machine Learning

This project leverages machine learning to detect and correct stuttering in real-time speech. Using advanced models like Wav2Vec 2.0, the system transcribes, detects, and corrects stuttered speech, offering a more inclusive communication experience. The project demonstrates a 95% detection accuracy with an F1 score of 0.95.

Features

• Stuttering Detection: Detect stuttered speech using a fine-tuned Wav2Vec 2.0 model.
• Stuttering Correction: Maps stuttered speech to fluent transcriptions.
• Real-Time Application: Supports live audio input and file uploads via an intuitive user interface.
• Data Augmentation: Techniques such as pitch shifting, time stretching, and additive noise are applied for robust model training.

Dataset Sources

• UCLASS Dataset: Annotated stuttered speech data.
• SEP-28k Dataset: A comprehensive dataset for stuttering event detection.
• C4 Grammar Error Correction Dataset: For fine-tuning the correction model.

Technology Stack

• Machine Learning: Wav2Vec 2.0 for speech-to-text and stutter detection.
• Natural Language Processing: Fine-tuned T5 model for grammatical corrections.
• Backend: Flask for handling requests.
• Frontend: User-friendly interface for audio input and transcription.

Team Contributions

Team Member	Role
Ajayvir Singh Sandhu	ML Model Development
Priyanshu Shukla	Backend and Deployment
Shubham Jha	NLP and Frontend Interface
Harasees Kaur	Data Preprocessing and Feature Extraction

Future Scope

• Incorporate voice output for corrected speech.
• Deploy the system on the cloud for multi-user scalability.
• Extend the dataset to support diverse accents and languages.

How to Run

1. Clone the repository:

   git clone https://github.com/your-repo-name.git
   cd your-repo-name

2. Install dependencies:

   pip install -r requirements.txt

3. Start the backend server:

   python app.py

4. Access the application through your browser at http://localhost:5000.

---