The psychological classification model of Transformer based on physiological signals
Aiming at the psychological classification and stress identification of college students,the multi-modal fusion and un-disturbed psychological classification algorithm based on Transformer architecture is studied and designed.A multiscale attention mod-ule is introduced to process facial expression and pulse wave features.Meanwhile,the Interaction-Attention mechanism was adopted in the feature fusion stage and synergistic higher order feature expression was introduced in the Transformer architecture.The results show that in the denoising processing of pulse wave,the average value of the difference based on Transformer architecture is 0.03,whose 95%confidence interval is-1.09 to 1.11,and the results of all test groups are within this range,which is significantly better than the traditional denoising method.At the same time,the average accuracy of the proposed algorithm is as high as 91.36%,which is 17.98%,9.22%and 6.79%higher than that compared with the other algorithms.Its average recall rate is as high as 88.51%,and its average F1 value is as high as 90.82%.It shows that the proposed algorithm has high accuracy and reliability in psychological classification and stress identification,which provides new ideas and methods for future psychological research and application.