Emotional-classification-through-voice-using-Backpropagation

Simple neural net to classify the emotion in an audio

Topic

In this notebook I experienced with audio classification, the dataset has 1400 audios of spoken sentences in each of the 7 emotional states (happy, angry, pleasant, disgust, sad, neutral and fear). My goal is to preprocess the raw audio into a form that a neural network can learn from, for that I used Melspectrograms, and them build a model that can accurately classify a voice tone as being in one of the mentioned class.

Objectives

• Process raw audio data to be neural net ready
• Build a classififer that can tell the emotion in a voice

Summary

• Extracting audio files
• Extracting Mel Spectrograms
• Audio transformation
• Creating the target dataset
• Train/Test split
• Creating the data loaders
• Building the Classifier
• Training and inference
• Trying out the model
• Conclusion

Libraries

• Pandas
• Nmupy
• Torchaudio
• Librosa
• Iphython
• Sklearn
• Glob

Datasource

https://www.kaggle.com/datasets/ejlok1/toronto-emotional-speech-set-tess?fbclid=IwAR181U1jADr4HCc591qJb61d0OjZXHZ5PV-iAi7sEnn-tLJQ2vaHTcce4Qk