Limitations of Emotion Recognition Methods in Usability Testing: A Case Study of Facial Expression Recognition on Smart Home Terminal Interfaces

Abstract

With increasing market demands for convenient and accurate emotional feedback, emotion recognition technology has become a preferred tool for evaluating the usability of interaction design, owing to its precision, stability, and high responsiveness in deriving emotional states from users' physiological indicators. However, the mapping mechanism between emotion recognition outputs and usability assessments remains unclear and under-defined. This study investigates the application of facial expression recognition technology in usability testing for smart home terminal interfaces, aiming to clarify the mapping characteristics between objective and subjective data in such contexts and to resolve erroneous correlations between emotional representations and usability judgments.First, we identify the relationship between trend-based emotional states and usability evaluations and propose a method to isolate effective instantaneous emotions. Second, we optimize the emotional calibration range and reveal the matching pattern between transient emotions and their designated calibration domains. Finally, through the fusion of dual-dimensional data, we correct recognition errors and propose a bidimensional feedback optimization method suitable for nonlinear mapping, which is further validated through experimental testing.This method effectively overcomes the limitations of traditional emotion recognition technologies in capturing subtle emotional fluctuations and filtering out irrelevant emotional responses, offering a new approach for enhancing the reliability of emotion-based usability evaluation.

Keywords: Emotion Recognition, Usability Testing, Facial Expression Analysis, Threshold Calibration, Affective Computing

DOI: 10.54941/ahfe1007523