Tourist Experiences Recommender System Based on Emotion Recognition with Wearable Data.

Area of Science:

• Computer Science
• Human-Computer Interaction
• Affective Computing

Background:

• Consumer-grade wearable devices enable widespread physiological data collection.
• Low-accuracy sensors offer potential for novel applications beyond healthcare.
• Real-world emotion recognition (ER) from physiological data remains challenging.

Purpose of the Study:

• Propose a tourist experiences recommender system (TERS) architecture.
• Develop an emotion recognition (ER) model using heart rate (HR) data from wearables in daily life.
• Integrate ER into a TERS to personalize tourist recommendations.

Main Methods:

• Collected HR measurements and user-labeled emotions via mobile applications.
• Employed deep learning algorithms, including hybrid Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks for ER.
• Designed and validated the Tourist Experience Recommender System - Emotion Recognition (TERS-ER) using Collaborative Filtering (CF) with CNN.

Main Results:

• Generated a dataset of HR measurements labeled with emotions from daily life.
• The CNN-LSTM hybrid model demonstrated promising performance for ER from HR data.
• Collaborative Filtering (CF) enhanced with CNN achieved superior performance for the TERS.

Conclusions:

• Emotion recognition from wearable HR data in real-world settings is feasible.
• Deep learning models, particularly CNN-LSTM, are effective for ER using HR data.
• The proposed TERS-ER system effectively integrates emotion recognition for personalized tourist recommendations.