Signal semantic recognition method based on multistage semantic representation and semantic calculation

Abstract

The invention provides a semantic recognition method based on multistage semantic representation and semantic calculation, and mainly solves the problems of poor interpretability and low generalization ability of signal semantic recognition in the prior art. The implementation scheme of the method comprises the steps of collecting a training set and a test set; constructing a signal semantic recognition network composed of cascaded multistage semantic representation networks and semantic computing networks so as to carry out learnable multistage semantic representation on the signals, and computing semantic categories of the signals according to the semantic representation; setting a semantic representation loss function and a cross entropy loss function to train the multistage semantic representation network and the semantic computing network in sequence to obtain a trained signal semantic recognition network; and obtaining a semantic recognition result of the to-be-recognized signalbased on the trained signal semantic recognition network. According to the method, the interpretability and generalization ability of signal semantic recognition are effectively improved. The method can be used for man-machine interaction and semantic information retrieval.

Description

technical field [0001] The invention belongs to the technical field of artificial intelligence, and in particular relates to a signal semantic recognition method, which can be used for human-computer interaction and semantic information retrieval. Background technique [0002] Signal semantic recognition refers to determining the semantic category to which it belongs according to the characteristics of the signal. [0003] Before the application of deep learning technology, researchers generally used methods such as bag-of-words feature matching or random forest for semantic recognition of signals. Based on the bag-of-words feature matching method, firstly, each semantic category is represented by a set of features obtained through feature engineering, and then the semantic category is judged by calculating the overall similarity between the sample signal and the feature set. The random forest first independently predicts the semantic category of the sample signal by multip...

AI Technical Summary

Problems solved by technology

[0005] First, since the discriminant features for recognition are obtained by directly adding the semantic features extracted by the semantic capsule network and the poorly interpretable image features extracted by the convolutional neural network, its interpretability is extremely poor;

[0006] Second, since the semantic capsule network uses fixed semantic primitives, it is necessary to design different semantic primitives for different problems before performing semantic feature extraction, and the entire network adopts an end-to-end training method, and the features between each semantic capsule Not independent, so it is difficult to transfer to other problems, resulting in insufficient generalization ability of the method

Images