Emotion Recognition from Physiological Signals

Overview

This project demonstrates emotion recognition using physiological signals like temperature, electrocardiogram (ECG), and galvanic skin response (GSR). It involves data preprocessing, model comparison, hyperparameter tuning, and deployment.

Dataset

The project uses a dataset named sensor_data.csv, containing physiological sensor readings and corresponding emotion labels. The dataset should have the following columns:

• Timestamp
• Temperature
• ECG
• GSR
• Emotion

Steps

1. Data Preprocessing:

   • Loads the dataset from sensor_data.csv.
   • Drops the Timestamp column.
   • Encodes emotion labels using Label Encoding.
   • Standardizes physiological features (Temperature, ECG, GSR) using StandardScaler.
   • Splits the data into training and testing sets.

2. Model Comparison and Automatic Selection:

   • Compares five classification models: Random Forest, SVM, KNN, Logistic Regression, and XGBoost.
   • Evaluates each model using accuracy, precision, recall, and F1-score.
   • Selects the best model based on the F1-score.

3. Hyperparameter Tuning of the Best Model:

   • Performs hyperparameter tuning on the selected best model using GridSearchCV.
   • Optimizes the model's performance by finding the best hyperparameter combination.

4. Final Model Training and Deployment:

   • Trains the final model with the best hyperparameters on the entire training set.
   • Evaluates the final model on the test set.
   • Saves the final model and scaler using joblib for deployment.

Dependencies

• Python 3.x
• pandas
• scikit-learn
• TensorFlow/Keras
• seaborn
• matplotlib
• joblib

Usage

1. Install the required dependencies.
2. Place the sensor_data.csv file in the same directory as the code.
3. Execute the code in a Jupyter Notebook or Python environment.
4. The trained model will be saved as final_emotion_model.pkl and the scaler as scaler.pkl.

Deployment

The saved model and scaler can be used to predict emotions from new physiological data. Load the model and scaler using joblib.load() and preprocess the new data using the same steps as in the training pipeline before making predictions.