Student behavior modeling and adaptive intervention system based on multi-modal perception and data fusion

Abstract

Disclosed is a student behaviour modelling and adaptive intervention system based on multi modal perception and data fusion. The system adopts a hardware-defined “terminal-edge-cloud” three-level distributed architecture and includes an edge perception and processing apparatus, a regional edge security gateway, a central fusion computing server cluster, a policy decision and distribution server, and a user interaction terminal. The edge apparatus converts an original audio and video into non-identifiable feature vectors locally and immediately overwrites original data; the security gateway encrypts and aggregates the features and assigns an untraceable temporary identifier; a central privacy computing accelerator card executes fusion computing on ciphertext data in a physically isolated memory; the decision server generates policy code through a hardware rule engine and performs distribution; and the user terminal parses an instruction into an anonymous prompt by using a local encrypted context.