🧠 AI/ML Augmentative and Alternative Communication (AAC) System

This project is a machine learning-powered Augmentative and Alternative Communication (AAC) system that enables users with speech or motor impairments to communicate using intelligent, context-aware speech suggestions and voice output.

🔍 Overview

The system captures text or speech input, interprets environmental data, detects user intent and context, and generates appropriate responses using LLMs. It is modular, extensible, and optimized for real-time assistive use.

🧱 System Architecture

The system is built around 5 core modules:

1. Input Module

   • Handles text or microphone input
   • Collects environment data (e.g., time, location)

2. Processing Module

   • Intent recognition
   • Context adaptation (social, work, general)
   • Predictive text generation (DistilGPT2)

3. Output Module

   • Postprocessing and cleaning of model output
   • Converts text to natural-sounding speech using TTS

4. Storage Module (Planned)

   • Profiles, preferences, and interaction history

5. UI Module (Planned)

   • Web interface for text input and response display

🛠 Technologies Used

• Python 3.10
• OpenAI Whisper (speech recognition)
• Hugging Face Transformers (DistilGPT2, DistilBERT)
• PyAudio, NumPy, SciPy
• noisereduce (real-time audio cleaning)

👤 Author

Yujun Ge