Millimeter wave indoor positioning method based on deep learning
An indoor positioning and deep learning technology, applied in neural learning methods, services based on location information, services based on specific environments, etc., can solve the problem of low positioning accuracy, achieve the effect of low equipment cost and easy promotion and use
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
Method used
Image
Examples
Embodiment 1
[0037] Such as figure 1 As shown, a millimeter-wave indoor positioning method based on deep learning includes the following steps:
[0038] S1: Collect the real position coordinates of the sampling points and the received signals from multiple wireless access points AP offline to obtain the training data set;
[0039] S2: Preprocessing the training data set to obtain the preprocessed training data set;
[0040] S3: The preprocessed received signal is used as the input of the deep neural network DNN model, the real position coordinates are the target output of the DNN model, and the DNN model is trained offline;
[0041] S4: Collect and receive signals online in real time, perform data preprocessing and input them into the trained DNN model, output real-time position coordinates, and complete indoor positioning of millimeter waves.
[0042] In the specific implementation process, the present invention makes full use of the received signals of multiple wireless access points A...
Embodiment 2
[0046] More specifically, on the basis of Example 1, the step S1 is specifically:
[0047] Deploy N APs in the indoor area that needs to be positioned. These APs work through time division multiplexing or frequency division multiplexing, and distinguish the signals sent by different APs by receiving time or frequency;
[0048] The sampling points are uniformly selected on the plane where the millimeter wave receiver is located to obtain the real position coordinates of the sampling points; the received signals from each AP are collected through the sampling points to complete the acquisition of the training data set.
[0049] More specifically, the step S2 is as follows: splitting the real and imaginary parts of the received signals from each AP and normalizing them, and collecting the received signals from the i-th sampling point at this position Signals processed by normalization and split real and imaginary parts of different APs (Y i,1 ,Y i,2 ,Λ,Y i,N ) and the location...
Embodiment 3
[0060] In order to more fully illustrate the beneficial effects of the present invention, the effectiveness and advancement of the present invention will be further described below in combination with simulation analysis and results.
[0061] In the simulation system, in such as Figure 4 20 x 15 x 6m as shown 3 1 AP is set up in the hotel lobby, and the training data points are evenly sampled on a plane with a height of 1.5m at a density of 10cm×10cm. The signal provides transmission services for a single user in a Single Input Single Output (SISO) manner. The number of subcarriers used by the system is 64, the length of the cyclic prefix is 16, and the modulation method is 4-Quadrature Amplitude Modulation (4-Quadrature Amplitude Modulation, 4-QAM). The learning rate of DNN is 0.003, and the activation function adopts ELU function, whose expression is
[0062] In the specific implementation process, in order to evaluate the positioning performance of the positioning s...
PUM
Login to View More Abstract
Description
Claims
Application Information
Login to View More