Abstract
Today, with the rapid development of economic level, people's esthetic requirements are also rising, they have a deeper emotional understanding of art, and the voice of their traditional art and culture is becoming higher. The study expects to explore the performance of advanced affective computing in the recognition and analysis of emotional features of Chinese paintings at the 13th National Exhibition of Fines Arts. Aiming at the problem of "semantic gap" in the emotion recognition task of images such as traditional Chinese painting, the study selects the AlexNet algorithm based on convolutional neural network (CNN), and further improves the AlexNet algorithm. Meanwhile, the study adds chi square test to solve the problems of data redundancy and noise in various modes such as Chinese painting. Moreover, the study designs a multimodal emotion recognition model of Chinese painting based on improved AlexNet neural network and chi square test. Finally, the performance of the model is verified by simulation with Chinese painting in the 13th National Exhibition of Fines Arts as the data source. The proposed algorithm is compared with Long Short-Term Memory (LSTM), CNN, Recurrent Neural Network (RNN), AlexNet, and Deep Neural Network (DNN) algorithms from the training set and test set, respectively, The emotion recognition accuracy of the proposed algorithm reaches 92.23 and 97.11% in the training set and test set, respectively, the training time is stable at about 54.97 s, and the test time is stable at about 23.74 s. In addition, the analysis of the acceleration efficiency of each algorithm shows that the improved AlexNet algorithm is suitable for processing a large amount of brain image data, and the acceleration ratio is also higher than other algorithms. And the efficiency in the test set scenario is slightly better than that in the training set scenario. On the premise of ensuring the error, the multimodal emotion recognition model of Chinese painting can achieve high accuracy and obvious acceleration effect. More importantly, the emotion recognition and analysis effect of traditional Chinese painting is the best, which can provide an experimental basis for the digital understanding and management of emotion of quintessence.
Introduction: Bridging the Semantic Gap in Art
Recognizing emotions in art is significantly more complex than standard object detection. A primary challenge in this field is the "semantic gap": the inherent difficulty in mapping low-level visual elements—such as line, color, and texture—to the complex, subjective emotional responses evoked by traditional Chinese paintings. While Deep Learning (DL) has revolutionized general image recognition, it faces hurdles in fine art due to data scarcity and domain mismatch. Features learned from natural images (like everyday objects) do not always transfer well to the abstract and symbolic nature of traditional brushwork, often leading to under-fitting or over-fitting.
Methodology: Improved AlexNet and Affective Computing
To address these challenges, this research developed a multimodal emotion recognition model integrating several advanced techniques. The core of the model is an Improved AlexNet architecture. The standard AlexNet was modified by advancing the "local normalization" step to occur after pooling rather than before. This structural change enhances generalization, weakens over-fitting, and accelerates convergence.
Furthermore, the model incorporates a Chi-Square test for feature selection to handle data redundancy and noise, ensuring the network focuses on the most "emotionally relevant" data points. To combat small sample sizes, a two-layer transfer learning strategy was employed to bridge the gap between universal visual features and specific artistic semantics. The training process was further optimized using Weighted Cross-Entropy (WCE) combined with Online Hard Example Mining (OHEM), forcing the network to prioritize "difficult" samples.
Data Source: The 13th National Exhibition of Fine Arts
The primary data source for this study was a collection of Chinese paintings from the Thirteenth National Exhibition of Fine Arts in China. This exhibition represents the pinnacle of contemporary Chinese artistic achievement, featuring works that blend traditional techniques with modern themes.
The dataset was split into a 7:3 ratio for training and testing. The proposed model was benchmarked against several standard algorithms, including LSTM, CNN, RNN, standard AlexNet, and DNN.
Key Results and Insights
The proposed model outperformed all baseline algorithms across several metrics:
- Accuracy: The model reached 92.23% on the training set and 97.11% on the test set, exceeding the next best-performing standard algorithm by at least 4.66%.
- Efficiency: Training time stabilized at approximately 54.97 seconds, with a testing time of 23.74 seconds. The "SpeedUp" ratio—the efficiency gain from parallel processing—was significantly higher for the improved AlexNet than for standard DNNs or RNNs.
- Error Reduction: The time delay error approached zero, a marked improvement over the 350-450ms delays typically seen in standard Deep Neural Networks.
A critical insight from the study is that Chinese painting pursues an artistic state of being "like and not like" objective reality. The model's success suggests that deep learning can capture this "quintessence"—the subtle emotional spirit of traditional culture.
Limitations and Future Work
Despite its high accuracy, the model faces limitations regarding stylistic diversity. Traditional Chinese painting encompasses a vast range of techniques, from meticulous Gongbi to freehand Shuimo. A model trained on one style may require further refinement to generalize across all categories. Future research will focus on refining mathematical models to better account for these stylistic variations and exploring the digital management of "national quintessence" to help the public better appreciate fine art.
Summary updated