As media consumption through electronic devices increases, there is growing interest in ASMR videos, known for inducing relaxation and positive emotional states. However, the effectiveness of ASMR varies depending on each video’s characteristics. This study identifies key sound and image features that evoke specific emotional responses. ASMR videos were categorized into two groups: high valence–low relaxation (HVLR) and low valence–high relaxation (LVHR). Subjective evaluations, along with physiological data such as electroencephalography (EEG) and heart rate variability (HRV), were collected from 31 participants to provide objective evidence of emotional and physiological responses. The results showed that both HVLR and LVHR videos can induce relaxation and positive emotions, but the intensity varies depending on the video’s characteristics. LVHR videos have sound frequencies between 50 and 500 Hz, brightness levels of 20 to 30%, and a higher ratio of green to blue. These videos led to 45% greater delta wave activity in the frontal lobe and a tenfold increase in HF HRV, indicating stronger relaxation. HVLR videos feature sound frequencies ranging from 500 to 10,000 Hz, brightness levels of 60 to 70%, and a higher ratio of yellow to green. These videos resulted in 1.2 times higher beta wave activity in the frontal lobe and an increase in LF HRV, indicating greater cognitive engagement and positive arousal. Participants’ subjective reports were consistent with these physiological responses, with LVHR videos evoking feelings of calmness and HVLR videos inducing more vibrant emotions. These findings provide a foundation for creating ASMR content with specific emotional outcomes and offer a framework for researchers to achieve consistent results. By defining sound and image characteristics along with emotional keywords, this study provides practical guidance for content creators and enhances user understanding of ASMR videos.

Facial expression recognition (FER) plays a pivotal role in affective computing and human–computer interaction by enabling machines to interpret human emotions. However, conventional FER models often overlook individual differences in emotional intelligence (EI), which may significantly influence how emotions are perceived and expressed. This study investigates the effect of EI on facial expression recognition accuracy within the valence–arousal space. Participants were divided into high and low EI groups based on a composite score derived from the Tromsø Social Intelligence Scale and performance-based emotion tasks. Five deep learning models (EfficientNetV2-L/S, MaxViT-B/T, and VGG16) were trained on the AffectNet dataset and evaluated using facial expression data collected from participants. Emotional states were predicted as continuous valence and arousal values, which were then mapped onto discrete emotion categories for interpretability. The results indicated that individuals with higher EI achieved significantly greater recognition accuracy, particularly for emotions requiring contextual understanding (e.g., anger, sadness, and happiness), while fear was better recognized by individuals with lower EI. These findings highlight the role of emotional intelligence in modulating FER performance and suggest that integrating EI-related features into valence–arousal-based models could enhance the adaptiveness of affective computing systems.

Facial expression recognition (FER) in naturalistic settings is constrained by label ambiguity and variability in stimulus–response alignment. Adopting a data-centric perspective, this study examined whether emotional intelligence (EI)-stratified training data influence FER performance by treating EI as a qualitative factor associated with affective data consistency. Naturally elicited facial expressions were collected in a controlled emotion induction experiment with subjective arousal and valence ratings. Using response-driven labeling, neutral ratings were retained as indicators of ambiguity. Participants were grouped into High and Low EI based on the alignment between subjective evaluations and outputs from a pretrained affect estimator. Identical binary classifiers for arousal and valence recognition were trained while varying only the training data composition and evaluated across baseline, unambiguous, and ambiguous test sets using independent training repetitions with repetition-level statistical aggregation. EI-stratified training was associated with statistically detectable, context-dependent performance differences: group effects were observed primarily under baseline conditions and, to a lesser extent, under ambiguous conditions, whereas no reliable differences emerged under unambiguous conditions. Pooled discrimination differences were modest, but item-level analyses identified significant differences in classification correctness in specific task–condition combinations. Comparable patterns were observed across alternative backbone architectures. These findings indicate that FER performance in naturalistic contexts is influenced not only by model architecture but also by the statistical structure and internal coherence of the training data, supporting EI-informed data selection in ambiguity-prone scenarios.