thierryrudaseswa/KinyaWhis

The kinyaWhisper is Kinyarwanda ASR # Kinyarwanda Voice Assistant 🤖 An intelligent voice assistant for Kinyarwanda language interaction, developed as part of the Intelligent Robotics course. 🌟 Key Features 🎙️ Kinyarwanda Speech Recognition: Built on KinyaWhisper, fine-tuned for 16kHz audio. 🧠 Smart Understanding: Uses fuzzy logic for flexible and context-aware question matching. 📢 Natural Kinyarwanda Responses: Speech output generated via Kinyarwanda TTS. 🔇 Clean Audio Input: Enhanced noise reduction ensures clearer speech detection. 🎚️ Accurate Detection: Employs Voice Activity Detection (VAD) for precise speech segmentation. 🔄 Repetition Filtering: Eliminates redundant transcriptions for smoother results. 📊 Insightful Analytics: Visualize matched queries and system behavior. 🌐 User-Friendly Interface: Seamless interaction through a web app powered by Gradio. 🛠️ Technology Stack AI Backbone: Hugging Face Transformers Audio Handling: Librosa + SoundFile Language Processing: FuzzyWuzzy & Python-Levenshtein for similarity scoring Frontend Interface: Gradio Performance Enhancements: WebRTC-based VAD & NoiseReduce for real-time optimization ## Installation 💻 ### Prerequisites - Python 3.12 - FFmpeg (audio processing): ```bash # Ubuntu/Debian sudo apt-get install ffmpeg # macOS brew install ffmpeg # Windows (via chocolatey) choco install ffmpeg ``` ## Quick Start 🚀 - Clone repository ```bash git clone github.com cd KinyarwandaVoiceAssistant ```# Create a virtual environment python -m venv .venv # Activate the virtual environment # For Linux/macOS: source .venv/bin/activate # For Windows: .venv\Scripts\activate pip install -r requirements.txt { "qa_pairs": [ { "question": "Mwaramuce neza?", "answer": "Mwaramutse! Amakuru yanyu?" } ], "default_response": "Vugurura ikibazo." } ## Example Interactions 🗣️ | Raw Transcription | Matched Question | System Response | |-------------------|-----------|--- …