DMFLC: Short Video Classification Based on Deep Multimodal Feature Fusion and Low Rank Representation

Abstract With the rise of short videos on social media platforms, short video recommendation technology has become an important research area in artificial intelligence and big data analysis. The unique features of short videos, including textual, image, and speech information, make multimodal content analysis crucial for accurate recommendation results. In this paper, we propose a deep learning-based classification algorithm that incorporates audio, visual, and text modalities for short video classification. The proposed algorithm aims to overcome the limitations of existing feature fusion methods by considering both common and private parts between different modal features. In this paper, we propose a classification algorithm based on deep multimodal feature fusion for short videos, and build a network based on audio modality, visual modality and text modality using the consistency and low-rank representation of multi-view features. The similarity loss function is used to explore the similarity of features extracted from different modalities by the public domain network, and the difference loss function is used to explore the difference of features extracted from the same modal private domain network and public domain network, and the classification loss is used to guide the classification of global video features. We evaluate the performance of the proposed algorithm on a public dataset and compare it with state-of-the-art approaches. The experimental results demonstrate that our algorithm achieves superior classification accuracy and outperforms existing approaches.