Speech Emotion Detection

This project detects emotions from speech using a neural network model. It includes scripts for recording audio, preprocessing the data, and evaluating the model.

Model Architecture

The model is a sequential neural network composed of several 1D convolutional layers, activation functions, dropout layers, a max-pooling layer, and a dense layer. Here is a detailed description of each layer and its configuration:

1. Conv1D Layer 1:

   • Name: conv1d_7
   • Input Shape: [None, 216, 1]
   • Filters: 128
   • Kernel Size: 5
   • Activation: linear
   • Padding: same
   • Initializer: VarianceScaling

2. Activation Layer 1:

   • Name: activation_8
   • Activation: relu

3. Conv1D Layer 2:

   • Name: conv1d_8
   • Filters: 128
   • Kernel Size: 5
   • Activation: linear
   • Padding: same
   • Initializer: VarianceScaling

4. Activation Layer 2:

   • Name: activation_9
   • Activation: relu

5. Dropout Layer 1:

   • Name: dropout_3
   • Rate: 0.1

6. MaxPooling1D Layer:

   • Name: max_pooling1d_2
   • Pool Size: 8
   • Strides: 8

7. Conv1D Layer 3:

   • Name: conv1d_9
   • Filters: 128
   • Kernel Size: 5
   • Activation: linear
   • Padding: same
   • Initializer: VarianceScaling

8. Activation Layer 3:

   • Name: activation_10
   • Activation: relu

9. Conv1D Layer 4:

   • Name: conv1d_10
   • Filters: 128
   • Kernel Size: 5
   • Activation: linear
   • Padding: same
   • Initializer: VarianceScaling

10. Activation Layer 4:

    • Name: activation_11
    • Activation: relu

11. Conv1D Layer 5:

    • Name: conv1d_11
    • Filters: 128
    • Kernel Size: 5
    • Activation: linear
    • Padding: same
    • Initializer: VarianceScaling

12. Activation Layer 5:

    • Name: activation_12
    • Activation: relu

13. Dropout Layer 2:

    • Name: dropout_4
    • Rate: 0.2

14. Conv1D Layer 6:

    • Name: conv1d_12
    • Filters: 128
    • Kernel Size: 5
    • Activation: linear
    • Padding: same
    • Initializer: VarianceScaling

15. Activation Layer 6:

    • Name: activation_13
    • Activation: relu

16. Flatten Layer:

    • Name: flatten_2

17. Dense Layer:

    • Name: dense_2
    • Units: 10
    • Activation: linear
    • Initializer: VarianceScaling

18. Activation Layer 7:

    • Name: activation_14
    • Activation: softmax

Setup

1. Clone the repository.
2. Install the dependencies using pip install -r requirements.txt.
3. Run the project using python main.py.

Usage

• preprocess.py: Contains functions for recording and preprocessing audio.
• evaluate.py: Contains the evaluation logic for the model.
• model.py: Loads the model architecture from a JSON file.
• utils.py: Contains utility functions and label encodings.
• main.py: Main entry point for the project.