Noisy Human Recognition

Recognized non-speech human sounds such as:

Clapping 👏 Footsteps 🦶 Brushing Teeth 🪥 Drinking Sipping 🧃 Laughing 😂 Breathing 🌬️ Crying Baby 😭 Coughing 🤧 Snoring 😴 Sneezing 🤧

Demo

Test the model at: link (hugging face space)

Datasets

ESC-50 dataset. Using only non-speech humman sounds.

Models

Classical ML Models:

• Random Forest
• SVM
• XGBoost

CNN-Based Models:

• CNN on Mel spectrogram
• CNN on Mel spectrogram + Deltas
• CNN on Mel spectrogram + Deltas + Augmentations
• CNN on Mel spectrogram + Deltas + Trimmed Audios
• CNN-LSTM