Abstract There was conducted an experimental comparison of the objective (FaceReader 4.0) and subjective assessment of the severity of emotions on different scales. As objects, 18 short (no more than 2 s) videos were used, on which the same person (actress) expressed the given emotions by pronouncing the word “Yes”. These stimuli were selected based on previously conducted subjective scaling experiments. We used (1) a direct subjective quantitative assessment of each video clip separately for each of the parameters selected in the FaceReader 4.0 program (neutral, happy, angry, disgust, fear, sad, surprise) – the severity of the corresponding emotion was evaluated in points; (2) a direct subjective assessment of each video clip for each of the parameters allocated in the 4-dimensional spherical model of emotions (bad – good; confidence – surprise; attraction; defensive reaction: fear – aggression); (3) the methodology of the Semantic Differential Scaling (SD) by C. Osgood; (4) the PXLab Self-Assessment-Manikin Scales. As a result, the relationship of all the scales used was investigated and it was shown that the objective qualification of emotions (using the FaceReader 4.0 program) does not coincide with the results of subjective assessment of the same samples.

中文翻译：

---

面部表情和主观情绪评估

摘要 对不同尺度上情绪严重程度的客观（FaceReader 4.0）和主观评估进行了实验比较。作为对象，使用了 18 个短视频（不超过 2 秒），其中同一个人（女演员）通过发音“是”来表达给定的情绪。这些刺激是根据先前进行的主观缩放实验选择的。我们对每个视频片段分别使用（1）直接主观量化评估，针对 FaceReader 4.0 程序中选择的每个参数（中性、快乐、愤怒、厌恶、恐惧、悲伤、惊讶）——评估相应情绪的严重程度点数；(2) 对分配在 4 维球形情绪模型中的每个参数的每个视频剪辑的直接主观评估（坏 - 好；信心 - 惊讶；吸引力；防御反应：恐惧 - 攻击性）；(3) C. Osgood 的语义差分缩放 (SD) 方法；(4) PXLab 自我评估-人体模型量表。结果，调查了所有使用的量表的关系，结果表明情绪的客观限定（使用FaceReader 4.0程序）与相同样本的主观评估结果不一致。