Application and Analysis of Deep Learning on Sleep Quality Prediction Tasks

With the progress of social economy and the acceleration of the pace of life, the problem of sleep quality has gradually become a global problem affecting human health. According to statistics from authoritative organizations such as the World Health Organization (WHO) and the American Academy of Sleep Medicine (AASM), more than 30% of adults have some degree of sleep quality problems, of which 10%-15% suffer from chronic memory loss. Traditional sleep quality assessment methods, such as Polysomnography (PSG) -based monitoring and questionnaire survey, are considered to be the routine standards for evaluating sleep quality, but PSG detection needs to be carried out in a professional sleep laboratory, which inevitably leads to its high cost and difficult to carry out long-term shortcomings. However, questionnaire survey is easy to be affected by individual stage mental state, and the accuracy is not high. Therefore, an accurate and objective method that can automatically assess sleep quality is urgently needed to help people achieve convenient and continuous sleep monitoring in their lives. This paper aims to systematically analyze the specific application status of deep learning models in sleep quality prediction tasks, discuss the advantages and disadvantages of different deep learning models in different sleep quality prediction tasks, analyze the limitations faced by current models, and propose some optimization strategies that may improve the performance of models in the future.