Handling concept drift in Wi-Fi-based localization
Systems and methods are provided for human activity analysis and localization using channel state information (CSI). A streamlined data processing and feature extraction approach that addresses concept drifts for time series data is provided. Data obtained from CSI or any other mechanism is used to estimate the wireless channel between two different wireless nodes (e.g., an access point (AP) and an associated station (STA)), and can be used to train a robust system capable of performing room level localization. A phase and magnitude augmented feature space along with a standardization technique that is little affected by drifts is also used.
1. A method for handling drift in wireless signals, the method comprising:
collecting channel state information through a transmitter and a receiver of a wireless access point;
preparing one or more data frames based on the channel state information, wherein preparing the data frames includes noise removal and normalization;
obtaining one or more statistical properties that are representative for a predefined area of interest; and
leveraging the one or more statistical properties to eliminate outliers from the prepared data frames.
2. The method of claim 1 , wherein the data frames are prepared along a time-axis.
3. The method of claim 2 , wherein the data frames are prepared along a frequency-axis.
4. The method of claim 3 , further comprising pre-selecting a fixed number of uniformly spaced frequency bands from a plurality of available subcarriers.
5. The method of claim 4 , further comprising calculating a score for performance evaluation, and using the score to compare the performance evaluation to performance of a baseline that assumes motion only affects the channel state information along the time-axis.
6. The method of claim 5 , wherein the score is a weighted average of precision and recall.
7. The method of claim 1 , further comprising measuring one or more sudden peaks in the channel state information and discarding the corresponding packets of the sudden peaks from the channel state information.
8. The method of claim 1 , further comprising extracting phase information from the channel state information, and removing a DC frequency component of the phase information from the channel state information using a Hampel filter.
9. A system for handling drift in wireless signals, the system comprising:
a wireless access point that collects channel state information through a transmitter and a receiver; and
a cloud server that:
prepares one or more data frames based on the channel state information, wherein preparing the data frames includes noise removal and normalization;
obtains one or more statistical properties that are representative for a predefined area of interest; and
leverages the one or more statistical properties to eliminate outliers from the prepared data frames.
10. The system of claim 9 , wherein the cloud server prepares the data frames along a time-axis.
11. The system of claim 10 , wherein the cloud server further prepares the data frames along a frequency-axis.
12. The system of claim 11 , wherein the cloud server further pre-selects a fixed number of uniformly spaced frequency bands from a plurality of available subcarriers.
13. The system of claim 12 , wherein the cloud server further calculates a score for performance evaluation, and uses the score to compare the performance evaluation to performance of a baseline that assumes motion only affects the channel state information along the time-axis.
14. The system of claim 13 , wherein the score is a weighted average of precision and recall.
15. The system of claim 9 , wherein the cloud server further measures one or more sudden peaks in the channel state information and discards the corresponding packets of the sudden peaks from the channel state information.
16. The system of claim 9 , wherein the cloud server further extracts phase information from the channel state information, and removes a DC frequency component of the phase information from the channel state information using a Hampel filter.
17. A non-transitory, computer-readable storage medium, having embodied thereon a program executable by a processor to perform a method for handling drift in wireless signals, the method comprising:
collecting channel state information through a transmitter and a receiver of a wireless access point;
preparing one or more data frames based on the channel state information, wherein preparing the data frames includes noise removal and normalization;
obtaining one or more statistical properties that are representative for a predefined area of interest; and
leveraging the one or more statistical properties to eliminate outliers from the prepared data frames.