The Affective Audio Dataset (AAD)

The Affective Audio Dataset (Ridley et. al, 2024) is a bespoke collection of 780 sounds of non-musical, non-anthropomorphic origin, derived from the BBC Sounds Library, annotated for evoked emotional features of Arousal and Valence. The dataset is intended for use in training and testing Machine Learning to predict the evoked emotion a sound induces in human listeners. No claim of copyright of the source sounds is made by the authors.情感音频数据集（Affective Audio Dataset，Ridley等人，2024年）是一套专属收录的780段非音乐、非拟人化声源音频，素材源自BBC声音库（BBC Sounds Library），并针对唤醒度（Arousal）与效价（Valence）两类诱发情绪特征完成标注。本数据集旨在用于训练与测试机器学习（Machine Learning）模型，以预测单段音频在人类听众中所诱发的情绪反应。作者未就该数据集的源音频主张任何版权。