Speech-Emotion-Analyzer

Why we included this project

For a hands-on look at how raw speech becomes classifier input, this project lays out the pipeline from start to finish. It pulls short clips from the RAVDESS and SAVEE datasets, extracts audio features with LibROSA, and feeds the resulting arrays into a convolutional network built with Keras. The model is trained to recognize five emotions, angry, calm, fearful, happy, and sad, in both male and female voices, and the README shows sample clips as waveforms and spectrograms so you can see what the model is actually working with. Treat it as a reference implementation for learning the mechanics rather than something to deploy as-is; the code is simple enough to fork, swap in your own recordings, or reshape for similar spoken-audio analysis.