Linear Predictive Coding for Acute Stress Prediction from Computer Mouse Movements.

Area of Science:

• Biomechanical modeling
• Human-computer interaction
• Physiological computing

Background:

• Prior research indicated Linear Predictive Coding (LPC) can approximate muscle properties from mouse movements for stress detection.
• A theoretical mass-spring-damper (MSD) model is often used to estimate arm biomechanics, including stiffness and damping.

Purpose of the Study:

• To validate Linear Predictive Coding (LPC) by comparing its derived muscle stiffness and damping parameters with those from a theoretical mass-spring-damper (MSD) model.
• To assess the efficacy of LPC and MSD models in predicting acute stress levels from computer mouse movements.

Main Methods:

• Linear Predictive Coding (LPC) was applied to computer mouse movement data to derive damping frequency and damping ratio.
• A theoretical mass-spring-damper (MSD) model was used to estimate corresponding biomechanical parameters.
• Correlation analysis was performed between LPC-derived and MSD-derived parameters.
• Binary classification was used to compare stress prediction performance between LPC, MSD, and neural network baselines.

Main Results:

• A significant positive correlation was found between damping frequency and damping ratio values derived from LPC and the theoretical MSD model.
• Stress level classification performance was comparable across LPC-based, MSD-based, and neural network-based methods.
• The efficacy of both LPC and MSD models for stress prediction was particularly evident with longer mouse movement trajectories.

Conclusions:

• The study validates the use of Linear Predictive Coding (LPC) as a reliable method for approximating muscle stiffness and damping from computer mouse movements.
• LPC offers a viable approach for predicting acute stress levels, demonstrating comparable accuracy to established biomechanical models.
• The findings support the integration of LPC-based biomechanical analysis in human-computer interaction for real-time stress monitoring.