Signal-strength-based deep learning regression fusion positioning method

Abstract

The invention, which relates to the field of signal-strength-based positioning, provides a signal-strength-based deep learning regression fusion positioning method. The method comprises: step one, establishing a fingerprint database in a to-be-located environment; step two, preprocessing data in the fingerprint database; step three, inputting the preprocessed data into an automatic coding model for pre-training; step four, constructing a DNN model based on the automatic coding model and training the DNN model; step five, constructing a CNN model and inputting preprocessed data into the CNN model for training; step six, according to a DS evidence fusion theory, carrying out fusion on output probability values of the DNN model and the CNN model and calculating a prediction result; step seven, according to an error value between a model estimation result and a real result, adjusting the DNN model and the CNN model; and step eight, according to the adjusted classification models, carryingout real-time positioning. According to the invention, features are extracted comprehensively by using complementary advantages of the DNN and CNN models, so that the accuracy of positioning is improved.

Description

technical field [0001] The invention relates to the field of positioning methods based on signal strength, in particular to a deep learning regression fusion positioning method based on signal strength. Background technique [0002] With the rapid development of the big data era, the demand for positioning technology is also growing rapidly, and its main driving force is the huge application and commercial potential that positioning services can bring. Through location-based services, users can determine their location in large supermarkets, find a product in a specific store, and find the facilities they need in public service places such as airports or train stations. The hospital confirms the location information of medical personnel or medical equipment, and the fire department locates the firefighters, etc. Therefore, positioning technology in complex environments is the core technology of several key issues of the Internet of Things. [0003] Literature M.Nowicki and...

AI Technical Summary

Problems solved by technology

[0004] The object of the present invention is to provide a deep learning regression fusion positioning method based on signal strength in order to solve the problem that the existence of the following three problems leads to inaccurate source location estimation results in complex environments: (1) classification model positioning The error is easily affected by the grid point spacing when the offline fingerprint database is established. (2) The feature extraction of the signal strength by the single deep learning model DNN is not comprehensive, and the signal strength distribution characteristics between the single training data cannot be learned. (3) If the model When locating grid points, the uncertainty is strong, directly using the label corresponding to the maximum value of the output layer as the positioning result will lead to a large positioning error

Images