Use VGGish Embeddings for Deep Learning in Simulink

This example shows how to use a simple neural network in Simulink® to classify audio signals from their VGGish feature embeddings using the VGGish Embeddings and Predict (Deep Learning Toolbox) blocks.

The network is a small fully connected network that was trained on VGGish feature embeddings extracted from air compressor audio signals. The air compressor data set consists of recordings from air compressors in a healthy state or one of seven faulty states. For information on how the network was trained, see Use VGGish Embeddings for Deep Learning.

While the simulation is running, you can change the input sound by double clicking the Select Compressor State block and choosing a type of sound from the drop-down menu. After you change the air compressor sound, see how the predicted class probabilities change.