Method and system of detecting dynamic status of a space

1. A detecting method for detecting a dynamic status in a space, wherein at least two wireless communication devices are deployed in the space and capable of performing a channel state detection to obtain a channel state information, the detecting method comprising:

controlling the at least two wireless communication devices to perform the channel state detection in a registration stage to obtain a plurality of registration-stage channel state information;

determining an environmental data of the space according to the plurality of registration-stage channel state information;

controlling the at least two wireless communication devices to perform the channel state detection in a detection stage to obtain a plurality of detection-stage channel state information; and

determining an intrusion situation of the space according to the environmental data and the plurality of detection-stage channel state information;

wherein the step of determining the environmental data of the space according to the plurality of registration-stage channel state information comprises:

using a principal component analysis to convert the plurality of registration-stage channel state information into a plurality of registration-stage principal component data; and

using a deep learning model to select a registration-stage principal component data from the plurality of registration-stage principal component data as the environmental data.

2. The detecting method of claim 1 , wherein a plurality of dynamic objects in the space have been moved out of the space in the registration stage.

3. The detecting method of claim 1 , wherein the step of converting the plurality of registration-stage channel state information into the plurality of registration-stage principal component data comprises excluding data with subcarrier amplitudes exceeding a threshold in the plurality of registration-stage channel state information.

4. The detecting method of claim 1 , wherein the step of using the deep learning model to select the registration-stage principal component data from the plurality of registration-stage principal component data as the environmental data comprises:

calculating a standard deviation of the plurality of registration-stage principal component data;

using the deep learning model to convert registration-stage principal component data with standard deviations less than a threshold in the plurality of registration-stage principal component data into a plurality of neural network embedded vectors, wherein an amount of the plurality of neural network embedded vectors is greater than a default value;

respectively calculating an average in-class degree of each neural network embedded vector of the plurality of neural network embedded vectors relative to other neural network embedded vectors; and

selecting the registration-stage principal component data corresponding to the neural network embedded vector with a largest average in-class degree in the plurality of neural network embedded vectors as the environmental data.

5. The detecting method of claim 1 , wherein the step of determining the intrusion situation of the space according to the environmental data and the plurality of detection-stage channel state information comprises:

using the principal component analysis to convert the plurality of detection-stage channel state information into a plurality of detection-stage principal component data;

using the deep learning model to analyze the environmental data and the plurality of detection-stage principal component data, to classify an environmental state of the space into a plurality of categories;

determining that there is at least one intruder in the space, when one of the plurality of detection-stage principal component data falls within an intrusion category of the plurality of categories.

6. The detecting method of claim 5 , wherein the step of converting the plurality of detection-stage channel state information into the plurality of detection-stage principal component data comprises excluding detection-stage channel state information with subcarrier amplitudes exceeding a threshold in the plurality of detection-stage channel state information, and converting the remaining detection-stage channel state information into the plurality of detection-stage principal component data.

7. The detecting method of claim 5 , further comprising generating a warning signal after determining that there is at least one intruder in the space.

8. The detecting method of claim 1 , further comprising training the deep learning model with the plurality of detection-stage channel state information.

9. A detecting system for detecting a dynamic status in a space, comprising:

at least two wireless communication devices, deployed in the space and capable of performing a channel state detection to obtain a channel state information; and

a computing device, comprising:

a processing unit, configured to execute a program code; and

a storage unit, coupled to the processing unit to store the program code, wherein the program code is configured to instruct the processing unit to execute the following steps:

controlling the at least two wireless communication devices to perform the channel state detection in a registration stage to obtain a plurality of registration-stage channel state information;

determining an environmental data of the space according to the plurality of registration-stage channel state information;

controlling the at least two wireless communication devices to perform the channel state detection in a detection stage to obtain a plurality of detection-stage channel state information; and

determining an intrusion situation of the space according to the environmental data and the plurality of detection-stage channel state information;

wherein the step of determining the environmental data of the space according to the plurality of registration-stage channel state information comprises:

using a principal component analysis to convert the plurality of registration-stage channel state information into a plurality of registration-stage principal component data; and

using a deep learning model to select a registration-stage principal component data from the plurality of registration-stage principal component data as the environmental data.

10. The detecting system of claim 9 , wherein a plurality of dynamic objects in the space have been moved out of the space in the registration stage.

11. The detecting system of claim 9 , wherein the step of converting the plurality of registration-stage channel state information into the plurality of registration-stage principal component data comprises excluding data with subcarrier amplitudes exceeding a threshold in the plurality of registration-stage channel state information.

12. The detecting system of claim 9 , wherein the step of using the deep learning model to select the registration-stage principal component data from the plurality of registration-stage principal component data as the environmental data comprises:

calculating a standard deviation of the plurality of registration-stage principal component data;

using the deep learning model to convert registration-stage principal component data with standard deviations less than a threshold in the plurality of registration-stage principal component data into a plurality of neural network embedded vectors, wherein an amount of the plurality of neural network embedded vectors is greater than a default value;

respectively calculating an average in-class degree of each neural network embedded vector of the plurality of neural network embedded vectors relative to other neural network embedded vectors; and

selecting the registration-stage principal component data corresponding to the neural network embedded vector with a largest average in-class degree in the plurality of neural network embedded vectors as the environmental data.

13. The detecting system in claim 9 , wherein the step of determining the intrusion situation of the space according to the environmental data and the plurality of detection-stage channel state information comprises:

using the principal component analysis to convert the plurality of detection-stage channel state information into a plurality of detection-stage principal component data;

using the deep learning model to analyze the environmental data and the plurality of detection-stage principal component data, to classify an environmental state of the space into a plurality of categories;

determining that there is at least one intruder in the space, when one of the plurality of detection-stage principal component data falls within an intrusion category of the plurality of categories.

14. The detecting system in claim 13 , wherein the step of converting the plurality of detection-stage channel state information into the plurality of detection-stage principal component data comprises excluding detection-stage channel state information with subcarrier amplitudes exceeding a threshold in the plurality of detection-stage channel state information, and converting the remaining detection-stage channel state information into the plurality of detection-stage principal component data.

15. The detecting system in claim 13 , wherein the program code further instructs the processing unit to generate a warning signal after determining that there is at least one intruder in the space.

16. The detecting system in claim 9 , wherein the program code further instructs the processing unit to train the deep learning model with the plurality of detection-stage channel state information.