AESI(Athens Emotional States Inventory)

开发生态有效的程序，用于收集可靠和无偏见的情绪数据，用于与针对精神障碍患者的社会和情感智能的计算机接口。随着雅典情绪状态清单 (AESI) 的发展，提出了针对五种情绪状态的视听数据库的设计、记录和验证：愤怒、恐惧、快乐、悲伤和中性。 AESI 的项目由句子组成，每个句子都有表示相应情绪的内容。情绪内容通过对 40 名年轻参与者的调查进行评估，问卷采用拉丁方设计。被 85% 的参与者正确识别的情感句子被记录在带有麦克风和摄像头的隔音房间中。 AESI 的初步验证是通过语音的自动情感识别实验进行的。生成的数据库包含 20 位母语人士用希腊语录制的 696 条话语，总持续时间约为 28 分钟。语音分类结果在 AESI 中自动识别情绪的准确率高达 75.15%。这些结果表明我们的方法在收集具有可靠内容、跨类别平衡和减少环境可变性的情绪数据方面是有用的。

We developed ecologically valid procedures to collect reliable and unbiased emotional data for computer interfaces targeting social and emotional intelligence support for patients with mental disorders. Along with the development of the Athens Emotional State Inventory (AESI), we present the design, recording, and validation of an audio-visual database covering five emotional states: anger, fear, happiness, sadness, and neutrality. The AESI corpus consists of sentences, each carrying content intended to convey a corresponding emotion. The emotional content of these sentences was evaluated via a survey administered to 40 young participants, adopting a Latin square design for the questionnaire. Emotional sentences correctly identified by 85% of the participants were recorded in a soundproof booth equipped with a microphone and a camera. Preliminary validation of the AESI database was conducted through automatic emotion recognition experiments on speech data. The finalized database contains 696 utterances recorded in Greek by 20 native Greek speakers, with a total duration of approximately 28 minutes. Speech classification experiments achieved an accuracy of up to 75.15% for automatic emotion recognition on the AESI database. These results demonstrate that our proposed method is effective for collecting emotional data with reliable content, balanced cross-category distribution, and minimized environmental variability.