R&D: Unobtrusive Mood Recognition System

Philips Research, Eindhoven

Project Details

Role: Lead Researcher & Primary Inventor
Methods: Experimental Design, Behavioral Coding, Quantitative Analysis (SPSS), Academic Writing
Outcome: Granted Patent, Published Paper (ACII 2011), and developed foundational algorithms for mood-sensing devices

the Challenge

Detecting Mood Without Wires

Philips Research wanted to explore the future of "Ambient Intelligence"—work environments that could detect a user's mood and adapt the atmosphere accordingly. The challenge was to determine the feasibility of a system that could recognize mood unobtrusively (i.e., without heart rate monitors or skin sensors). Existing studies were flawed: they relied on actors portraying exaggerated emotions for 2-3 seconds. My goal was to determine if we could detect mood based on natural body posture over longer periods (8 minutes) in a realistic environment.

Statistically Significant Signals: Data visualization showing the breakdown of "Head Tapping" frequency. My analysis revealed this specific behavior was a unique identifier for "Positive High" mood states, forming the basis of our patent.

My role

Experimental Design & Behavioral Coding

I designed the research protocol to bridge the gap between psychological theory and engineering constraints.

Metric Definition: Since this was a novel field, I had to determine what to measure. I developed and validated a rigorous coding system to track temporal changes across six postural dimensions (Head, Shoulders, Trunk, Arms, etc.).
Behavioral Analysis: I analyzed video data of users under mood-inducing conditions, coding naturalistic displays of posture.
Quantitative Validation: I used SPSS to analyze the frequency and duration of specific movements to find statistically significant correlations between posture and emotional state.

the outcome

A Patent on "Headbanging"

The research yielded a breakthrough discovery: specific head movements ("head tapping") were exclusively displayed during "Positive High" energy mood states, making the head the most reliable region for differentiation.

Intellectual Property: I filed and was granted a Patent as the main inventor. This patent outlined the algorithmic rules for designing products that detect mood via head movement.
Academic Recognition: The results were published in the Affective Computing and Intelligent Interactions conference (2011).
Feasibility Confirmed: The study successfully proved that natural postures could be used to train powerful recognition algorithms for consumer electronics.

Mood Recognition Based on Upper Body Posture and Movement Features