This repository is a voice activity detection (VAD) toolkit that implements multiple models (DNN, bDNN, LSTM, ACAM) for detecting speech versus non-speech in audio. It also provides a recorded dataset in varied real-world settings (e.g. bus stop, construction site, park, room) with ground truth labeling. Acoustic feature extraction (multi-resolution cochleagram, MRCG). Post-processing modules (e.g. smoothing, thresholds). The toolkit supports both MATLAB and Python/TensorFlow components (for feature extraction, classification, postprocessing). Acoustic feature extraction (multi-resolution cochleagram, MRCG). Provided real-world dataset with manual annotations.
Features
- Multi-model VAD: DNN, boosted DNN (bDNN), LSTM, ACAM (adaptive context attention model)
- Acoustic feature extraction (multi-resolution cochleagram, MRCG)
- Training scripts (MATLAB / Python) and pretrained models
- Post-processing modules (e.g. smoothing, thresholds)
- Provided real-world dataset with manual annotations
- Combined MATLAB + TensorFlow interoperability
Categories
Sound/AudioFollow VAD
Other Useful Business Software
Build Agents and Models on One Platform
Gemini Enterprise Agent Platform is Google Cloud's comprehensive platform for developers to build, scale, govern, and optimize agents and models. Choose from Google's most advanced models and third-party models like Anthropic's Claude Model Family.
Rate This Project
Login To Rate This Project
User Reviews
Be the first to post a review of VAD!