Research on modeling and recognition algorithm of human-computer interaction behavior in cognitive degeneration process of Alzheimer's disease

Alzheimer's Disease (AD) is a common neurodegenerative disease, which brings a heavy burden to patients and their families. Traditional diagnostic methods have limitations, such as high cost, strong invasiveness and strong subjectivity. With the development of human-computer interaction (HCI) technology, it is expected to realize the early diagnosis of AD and the quantitative evaluation of cognitive degradation process through multimodal data acquisition and intelligent analysis. In this paper, a modeling and recognition algorithm based on HCI behavior is proposed. By designing virtual daily tasks, the behavior performance of patients with cognitive impairment is systematically quantified, multimodal behavior data is extracted, and a cognitive state-behavior response function model is constructed. Furthermore, Hidden Markov Model (HMM) and Radial Basis Function Network (RBFN) are used to build the model, and the hybrid model structure of space-time graph convolution network (ST-GCN) and Long Short-Term Memory (LSTM) is combined to realize accurate recognition of cognitive degradation stage. The experimental results show that the model is superior in multi-stage cognitive degradation recognition, with an accuracy of 0.91 and good temporal resolution and classification accuracy. In addition, through the analysis of SHAP value, the model can capture the explanatory behavioral signs closely related to cognitive degradation, which provides a scientific basis for early intervention and disease tracking.