KanishkNavale/Speech-Emotion-Recognition

A simple CNN-LSTM deep neural model using Tensorflow to classify emotions from a speech dataset

Speech-Emotion-Recognition

A simple CNN-LSTM deep neural model using Tensorflow to classify emotions from speech dataset

Table of contents

• Reading the 'train.json' file
  • Download link: www.kaggle.com/dataset/1ef5f65351c643f6d2fb0becc9993b73be3130090e234fa88f9307ef18b9de78
• Rescaling the features
• Creating train & validation data splits
• Defining the DNN model
• Model training
• Evaluating Results

Overview

• Training Log,

  

• Results,

  

Dependencies

Install dependencies using:

pip3 install -r requirements.txt 

System Requirements

• The notebook is executed online on kaggle.com with a GPU accelerator.
• To run it locally on a PC, 16GB of RAM is needed.

Challenge

• Boost the 'Validation Accuracy' by Model modification or Data Augmentation

Contact

• email: navalekanishk@gmail.com