An interpretable method and device for signal modulation type recognition model based on neuron activation

Abstract

The invention discloses an interpretable method and device for a signal modulation type recognition model based on neuron activation. For large neurons, the feature matrix can be obtained by adding the activation values ​​of these neurons and taking the derivation of the input. The feature matrix reflects the model's attention to each part of the input. After the feature matrix is ​​printed on the input waveform diagram and the constellation diagram, the degree of attention of the model to each part of the input and the degree of attention of the model to the key positions on the constellation diagram can be obtained, and the interpretive analysis of the model can be realized. This method can interpretively analyze the recognition mechanism of the LSTM network-based signal modulation type recognition model, thereby verifying the security of the model.

Description

technical field [0001] The invention belongs to the field of deep learning security, and in particular relates to an interpretable method and device for a signal modulation type recognition model based on neuron activation. Background technique [0002] Wireless communication has played an increasingly important role in people's production and life, and has been used in many fields, such as in the civil field, through wireless communication to achieve vehicle-machine interconnection, vehicle automatic driving, road camera wireless monitoring, mobile phone Positioning and navigation, etc.; in the military field, the internal wireless interconnection of weapon platforms is realized through wireless communication, forming a short-range communication command network, precise positioning, etc. Wireless communication is implemented depending on the wireless communication system. The wireless communication system consists of a transmitter and a receiver. The transmitter performs c...

AI Technical Summary

Problems solved by technology

[0006] Aiming at the problems that the current signal modulation type identification model based on LSTM network is difficult to perform interpretability analysis and the safety of the model is difficult to be verified, the present invention provides an interpretable method and device for a signal modulation type identification model based on neuron activation , the feature matrix of the model pair input is obtained through the activation value of the neuron, and the mechanism of the LSTM model recognition signal is interpreted through the feature matrix to verify the security of the model

Images