Asian Journal of Mathematics and Computer Research

Aims: This study aims to construct an AI-based emotion-color mapping model with high precision, develop a lightweight and user-friendly mobile application (ArtThera) for college students' emotion regulation, and systematically verify its effectiveness in campus mental health intervention through rigorous randomized controlled trials. 

Study Design: A randomized controlled trial (RCT) integrating multi-methodological approaches, including systematic literature review, data-driven analysis with machine learning algorithms, iterative prototype development based on user-centered design, and comprehensive user testing combining physiological indicators and psychological scales. 

Place and Duration of Study: Conducted at the School of Digital Media and University Psychological Counseling Center of Chongqing Polytechnic University of Electronic Technology, spanning from June 2024 to June 2025 (12 months), including 6 months of model construction and APP development, and 6 months of user recruitment, intervention implementation, and data collection. 

Methodology: 1) Literature research: Systematically review classic and cutting-edge theories of color psychology (e.g., PAD emotional state model, HSV color space theory) and evidence-based research on art therapy, summarizing the limitations of current research and identifying research gaps; 2) Data-driven analysis: Use Python programming language to collect 1000+ emotion-labeled artworks from authoritative open databases (WikiArt, COSMOS), where WikiArt covers works by 500+ international artists and COSMOS provides standardized emotion annotation data. Extract color features (RGB values, brightness, saturation) using Pandas, NumPy, and OpenCV libraries, convert RGB values to HSV color space for standardized calculation (H=arctan2(√3(G-B), 2R-G-B), S=(max(R,G,B)-min(R,G,B))/max(R,G,B), V=max(R,G,B)/255), exclude abnormal data (e.g., color distortion) using the IQR method (calculate quartiles Q1 and Q3, define outliers as values less than Q1-1.5IQR or greater than Q3+1.5IQR), and apply K-means clustering (Scikit-learn library, k=8, init='k-means++', n_init=10, max_iter=500, convergence threshold=0.001) to group color features by 8 core emotion labels, establishing a scientific and quantifiable emotion-color rule library; 3) Prototype iteration: Develop the ArtThera APP using Flutter (frontend) and Firebase (backend), integrating multiple functions such as detailed emotion input, personalized color scheme generation, dynamic painting templates, white noise synthesis, vibration feedback, and "emotion memory capsules". Referring to the user-centered development approach of the LiveWell app, conduct semi-structured interviews with 30 college students first, then complete three rounds of prototype iteration and user testing; 4) User testing: Recruit 200 college students (100 males, 100 females; age 18-22 years) via the University Psychological Association, including 50 students with mild anxiety (SAS score 50-59). Divide participants into experimental group (using ArtThera) and control group (engaging in traditional static coloring) through random assignment. Evaluate intervention effects at three time points (T0: pre-intervention 1 day, T1: mid-intervention 2 weeks, T2: post-intervention 4 weeks) using Heart Rate Variability (HRV, including SDNN and RMSSD indicators), Self-Rating Anxiety Scale (SAS), and PANAS scale, combined with semi-structured interviews to collect user feedback. 

Results: 1) Emotion-color mapping model: Clear and significant correlations between emotions and color features were identified (e.g., anxiety→low brightness [35±5], low saturation [28±4], hue range 210-240°; excitement→high brightness [82±6], high saturation [75±7], hue range 30-60°), which are highly consistent with findings from classic mandala coloring studies; 2) APP performance: Fully meets preset technical indicators (iOS average launch time 1.2s, Android average 1.4s, both ≤1.5s; color scheme generation average response time 1.7s, dynamic template loading average 1.9s, both ≤2s), supporting iOS 12.0+ and Android 8.0+ systems. In preliminary testing (n=20), 85% of participants rated "ease of use" ≥4/5, and 90% reported that dynamic templates are "more engaging" than static coloring books, exceeding the average acceptance rate of adolescent mental health apps (78%); 3) Intervention effects: Expected to achieve ≥10% increase in HRV (SDNN indicator) in the experimental group, ≥12% decrease in SAS scores, and ≥15% increase in PANAS positive affect scores, which is comparable to or even better than mindfulness-based interventions. 

Conclusion: The AI-driven ArtThera system effectively addresses the limitations of traditional art therapy (lack of personalization, poor accessibility) and existing mental health apps (single function, insufficient integration of art therapy and data-driven technology), providing a scientific, convenient, and scalable tool for campus mental health services. It enriches the theoretical framework of digital art therapy and contributes to the development of data-driven psychological intervention strategies. 

Fund Project: 2025 Student Science and Technology Innovation "Mayue" Project of Chongqing Polytechnic University of Electronic Technology.