融合面部表情和社交网络的音乐情境推荐

Incorporating Facial Expressions and Social Networks for Musical Contextual Recommendations

摘要: 在数字音乐快速发展的同时，人们也从音乐匮乏的时代走向音乐过载的时代，越来越多的用户在使用数字音乐时，希望从种类繁多的音乐中得到其当前喜欢的推荐。不同于其他商品推荐，音乐的消费使用与用户的情绪高度关联。针对这些问题，提出了一种融合面部表情和社交网络的音乐推荐方法。该方法利用社交网络历史行为分析建立用户知识模型，再利用目前的面部识别技术，结合用户的面部表情去判断用户当前的情绪。在现有的音乐推荐方法基础上，加入或过滤出情绪分类以及与用户情绪符合的推荐结果。实验结果证明了情绪因素会对音乐偏好产生影响，提出的融合面部表情和社交网络的音乐情境推荐方法，比其他不考虑用户情绪单单基于内容推荐或者协同过滤推荐的推荐方法略好且稳定，具有较好的实际应用效果。

Abstract: At the same time as the rapid development of digital music, people are moving from the era of lack of music to the era of music overload. More and more users are hoping to get their current favorite recommendations from a wide variety of music when using digital music. Unlike other product recommendations, the consumption of music is highly correlated with the user’s mood. In response to these problems, a music recommendation method combining facial expressions and social networks is proposed. The method uses social network historical behavior analysis to establish a user knowledge model, and then uses the current facial recognition technology to combine the user’s facial expression to determine the user’s current mood. Based on the existing music recommendation method, the emotion classification and the recommendation result in accordance with the user’s emotion are added or filtered. The experimental results show that emotional factors have an impact on music preferences. The proposed music context recommendation method combining facial expressions and social networks is slightly better and more stable than other recommendation methods based on content recommendation or collaborative filtering. It has a better practical application result.