Multimodal Safety Detection System for Parkinson's Disease Based on Multi-head Attention Mechanism
During the actual remote diagnosis process of Parkinson's disease,it has a high misdiagnosis rate that the single-mode data is applied to detect Parkinson's disease,and the remote diagnosis has prominent security.To improve the accuracy and security of remote diagnosis of Parkinson's disease,a multimodal secure remote auxiliary detection system with a privacy protection function was designed.The dual-modal data of Parkinson's disease speech and gait were used to integrate the multi-head attention mechanism with multi-layer perceptron after traditional convolutional neural networks,which effectively improved the feature extraction,fusion,and recognition ability of the model.To ensure the security of the data transmission process,the differential privacy and noise method based on cosine chaos is used to disturb the randomly separated data numbers to improve the data transmission security of Parkinson's disease.The two-mode ablation experiment and comparison experiment results show that the actual test accuracy of the proposed multi-mode remote detection model of Parkinson's disease based on the multi-head attention mechanism reaches by 0.913.The evalua-tion indicators and convergence speed of the model are higher than that of the traditional model,which has a good effect on intelligent-ly assisted detection of Parkinson's disease.It can meet the needs of early innovative safety screening and diagnosis of Parkinson's disease.
万方数据
维普
