Mood-based On-Car Music Recommendations

Abstract: Driving and music listening are two inseparable everyday activities for millions of people today in the world. Considering the high correlation between music, mood and driving comfort and safety, it makes sense to use appropriate and intelligent music recommendations based on the mood of drivers and songs in the context of car driving. The objective of this paper is to present the project of a contextual mood-based music recommender system capable of regulating the driver's mood and trying to have a positive influence on her driving behaviour. Here we present the proof of concept of the system and describe the techniques and technologies that are part of it. Further possible future improvements on each of the building blocks are also presented.

• The paper introduces a mood-based system that integrates physiological, musical, and telematics data to shape in-car music recommendations.
• It employs wearable sensors for mood detection and social tag classifiers for categorizing tracks into emotional states.
• The system dynamically adapts music playlists using contextual factors like time, location, and driving behavior to improve safety and driver satisfaction.

Mood-based On-Car Music Recommendations

Introduction

The paper "Mood-based On-Car Music Recommendations" introduces a system that integrates mood recognition with music recommendation to enhance the driving experience. Recognizing the critical connection between music, mood, and driving behaviors, the system aims to adjust the driver’s mood through curated music playlists, thereby influencing driving comfort and safety.

System Overview

The proposed system functions by collecting a diverse range of data inputs to provide mood-aligned music recommendations. The architecture involves several key modules:

1. User Mood Detection: Captures physiological data from wearable sensors to infer driver mood based on heart rate dynamics.
2. Music Mood Recognition: Classifies musical tracks into four categories—Happy, Tender, Sad, and Angry—using a tag-based folksonomy, aligning with Russell’s circumplex model of affect.
3. Driving Style Recognition: Monitors telematics from OBD-II devices to characterize driver behavior as aggressive or calm.
4. Recommendation System: Utilizes contextual data such as time of day, location, and user preferences to generate appropriate music playlists.
5. Mobile Application: An Android-based interface for interaction with the recommendation system, compatible with both standalone and Android Auto configurations.

The integration relies on combining user mood, musical preferences, and technical driving data to foster a mood-conducive driving environment.

Technical Components

User Mood Detection leverages heart rate variability metrics to assess arousal and valence, fitting these into a framework that aligns with existing emotion recognition paradigms like Russell’s circumplex model. This methodology benefits from personalized calibration, circumventing the need for broad population data.

Music Mood Recognition employs classifiers built upon social tags derived from platforms such as last.fm, categorizing songs into mood labels that guide the recommendation process. The choice of categories reflects a balance between simplicity and the emotional granularity provided by more complex models.

Driving Style Recognition utilizes telemetric data captured via OBD-II interfaces or smartphone gyroscopic capabilities to determine driving aggression. While OBD-II offers finer resolution, smartphone-based approximations fill the gap where such hardware is unavailable.

Music Recommender System employs a contextually-aware model, where parameters like time and location dynamically influence playlist generation. The recommender system seeks to either maintain or modify the current mood depending on the desired driving outcome, be it calm and safe or attentive and alert.

Mobile Application serves as the user interface, featuring simplified controls to minimize driver distraction. It presents the application of recommendations through in-car systems, bridging the interface seamlessly with vehicle infotainment platforms utilizing Android Auto or proprietary solutions.

Practical Implications

The robust interaction between mood recognition and music recommendation presents transformative potential for in-car environments. Implementing such a system carries the promise of improving road safety and enhancing driver satisfaction. The architectural integration demonstrates a comprehensive use of contextual data, though expanding mood detection through additional modalities (e.g., facial recognition and skin conductance) could refine accuracy and responsiveness.

Conclusion

This work outlines a significant stride towards leveraging intelligent systems within automotive contexts for mood-influenced music recommendation. Future developments proposed include expanding mood categories, incorporating multiple detection modalities, and refining context-driven recommendation algorithms. This continuous evolution holds potential for further enhancing both safety and enjoyment for drivers, potentially setting a standard for future in-car infotainment systems.