A SUPERVISED ONLINE CROWD ANALYSIS NETWORK WITH DUAL-TASK DEEP LEARNING IDEA IN HEALTHCARE APPLICATION

As the most active research branch in the field of artificial intelligence, deep learning has gradually become a research hotspot in the medical field in recent years, and has achieved some success in medical image and signal processing, biomedical and healthcare, medical information mining and retrieval, showing great application prospects. Aiming at the problem of the uneven distribution of data in biomedical and healthcare area under smart city, this paper proposes a novel supervised network by a dual-task deep learning scheme. First, we optimize the early distribution by supervising the features of the early stage, which helps the shallow network to determine the distribution of data, and avoids the gradient disappearance and slow convergence. Then, a novel supervised network framework is proposed to focus on the prediction accuracy and generate crowd response maps that reflect the importance of different features. The final distribution revising module uses high-frequency semantic information to further supervise the predicted distribution. The noise data suppression module is proposed to suppress the misjudgment on fake biomedical data. Experimental results on four commonly healthcare used datasets show that our proposed deep learning model performs better in the biomedical data analysis and density distribution, and outperforms the state-of-the-art methods.