Feature Extraction of Ancient Chinese Characters Based on Deep Convolution Neural Network and Big Data Analysis

Area of Science:

• Computer Science
• Artificial Intelligence
• Digital Humanities

Background:

• Deep learning advancements have led to applications in various fields like image processing.
• Deep convolution neural networks (CNNs) are gaining attention for their potential in complex pattern recognition.
• Extracting ancient Chinese characters presents challenges due to variations in style and degradation.

Purpose of the Study:

• To explore the effectiveness of deep convolution neural networks for ancient Chinese character extraction.
• To compare the performance of deep CNNs against traditional machine learning algorithms for this task.
• To provide a robust method for identifying historical Chinese inscriptions.

Main Methods:

• The study details the structure model, pooling process, and network training of a deep CNN.
• A comparative analysis was conducted between the proposed deep CNN algorithm and traditional machine learning algorithms.
• The methodology involved training and testing the model on ancient Chinese characters from Ming Dynasty plaques.

Main Results:

• The deep CNN achieved peak accuracy and recall rates of 81.38% and 81.31% for Ming Dynasty plaque characters.
• With 50 training samples, the algorithm reached a recognition rate of 99.72%, outperforming other methods.
• The deep CNN demonstrated excellent performance across different dynasties, sample sizes, and interference factors.

Conclusions:

• Deep convolution neural networks combined with big data analysis offer a powerful solution for ancient Chinese character recognition.
• The proposed algorithm provides a reliable reference for historical inscription extraction and analysis.
• This approach has the potential to significantly advance the field of digital humanities and historical research.