This invention relates to a multimodal
perception and AI-driven emotion regulation method, comprising: simultaneously collecting multi-dimensional environmental parameters of the user and the user's own multimodal physiological signals; preprocessing the multi-dimensional environmental parameters and the physiological signals respectively, and extracting
feature data to generate corresponding environmental
feature data and physiological
feature data; based on the environmental feature data and physiological feature data, identifying the user's current emotional state and corresponding emotional intensity through a preset multimodal
emotion recognition model; when the user is identified to be in a preset negative emotional state, matching a combined intervention scheme corresponding to the current emotional state and intensity according to a preset intervention strategy
library, generating and executing a combined intervention instruction; the combined intervention instruction controls at least two different types of actuators to regulate the user through at least two sensory channels.