AUTHOR=Li Jing , Chen Dongliang , Yu Ning , Zhao Ziping , Lv Zhihan 

TITLE=Emotion Recognition of Chinese Paintings at the Thirteenth National Exhibition of Fines Arts in China Based on Advanced Affective Computing

JOURNAL=Frontiers in Psychology

VOLUME=Volume 12 - 2021

YEAR=2021

URL=https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2021.741665

DOI=10.3389/fpsyg.2021.741665

ISSN=1664-1078

ABSTRACT=The present work aims to investigate the performance of Advanced Affective Computing in recognizing and analyzing human affects in Chinese paintings at the Thirteenth National Exhibition of Fines Arts in China. A multimodal emotion recognition algorithm for Chinese paintings is designed based on improved AlexNet and Chi Square Test. In particular, AlexNet solves the “semantic gaps” in recognizing emotions from images Chinese paintings. Data redundancy and data noise in each modality of Chinese paintings are removed drawing upon Chi Square Test added to the improved AlexNet. At last, simulation experiments are designed to validate the proposed algorithm’s performance, in which the image data of Chinese paintings at the Thirteenth National Exhibition of Fines Arts in China are employed. Algorithms included for comparative simulations are LSTM (Long Short-Term Memory), CNN (Convolutional Neural Network), RNN (Recurrent Neural Network), AlexNet, and DNN (Deep Neural Network). Results demonstrate that the proposed algorithm can provide an emotion recognition accuracy of 92.23% and 97.11% on the training and the test datasets, respectively. As for time durations, it requires a training duration of about 54.97 seconds and a test duration of about 23.74 seconds. Regarding acceleration efficiency, it suits for processing massive amounts of image data and performs better on test dataset than training dataset, presenting a higher speedup ratio than other algorithms. Through experiments, the proposed algorithm is proved to be accurate and fast in recognizing and analyzing human affects in Chinese paintings, providing an experimental basis for the digital emotion interpretation and management of Chinese paintings.