Mood Detection in Aesthetically Appealing Video Based on Color Association

Aesthetic comes under the branch of cognitive science, and it plays a significant role in demonstrating the psychology of humans. Videos or images that are more colourful attracts lots of audiences. At the same time, these colour characteristic helps in finding the mood of a person whether they are happy, sad, anger, fear etc. Many approaches such as machine learning technique, neural network were designed earlier for achieving mood detection in appealing video. However, effective detection with enhanced accuracy was not attained in conventional methods. For overcoming these drawbacks, Deep Convolutional Neural Network (DCNN) based approach is designed in this proposed work to perform aesthetic classification and mood detection. At the beginning of the process video is converted into frames and key frame extraction is performed using histogram method. Then, foreground portion is separated from the background using object detection technique based on Gaussian Mixture Model (GMM). After that, extraction of low level, high level features and aesthetic classification is done using DCNN. Finally, using pleasing video the mood detection is done based on color features. The proposed method is implemented and its performances are evaluated using measures including accuracy, precision, recall, and F1 score whose values are 77.4, 77.4 77.9 and 77.4%. Based on this proposed approach automated and effective aesthetic classification can be achieved along with human mood detection.