Wifi based Indoor Line of Sight Identification

By: Nidhi Kumari

LINE OF SIGHT & NON-LINE OF SIGHT

IMPORTANCE OF LOS IDENTIFICATION

IMPORTANCE

Increasing the performance

Ensuring tight electromagnetic

coupling

Reducing the error

Optimizing our communication

system

CHALLENGES IN LOS IDENTIFICATION Physical layer information is yet

unexplored and wifi devices only report single-valued MAC layer RSS to upper layers.

Real-world evaluation of LOS identification is still a difficult task.

It is a labor intensive process as information need to be extracted from MAC layer RSS.

INTRODUCTIONIn this indoor LOS identification technique we exploit channel state information (CSI) from the PHY layer to identify the availability of LOS component indoors .

Use of natural mobility to magnify the distinction between LOS and NLOS conditions.

Use of frequency diversity to reveal the spatial disturbance of NLOS propagation.

MEASUREMENT AND INFORMATION EXTRACTION Channel State Information (CSI) : In this

LOS identification scheme we explore the recently available PHY layer information. The sampled version of channel frequency response (CFR) is revealed to upper layers in form of channel state information (CSI).

Measurements with CSI: (a) Shape-Based Features with CSI:

(b) Statistics-Based Features with CSI: signals travelling alog NLOS paths tend to behave more randomly compared with those along a clear LOS path Rician-k Factor

Channel Statistics with Mobility: Shape based features are infeasible due to insufficient bandwidth of WiFi.The main hurdle is short transmission distance which makes the NLOS path not adequately random,thus increasing difficulty.

LOS IDENTIFICATION

Preprocessing: Since lack of time and frequency synchronization induces phase noise when measuring channel response we revise the phase to reduce error.

Normalization: To make LOS identification independent of power attenuation we normalize CIR samples and CSI amplitudes by dividing them by average amplitude.

Identification: For a set of normalized CIR and CSI amplitude from N packets,

skewness feature hypothesis test is: H0 : s < sth H1 : s > sth

kurtosis feature hypothesis test is: H0 : κ > κth H1 : κ < κth

sth and κth are the identification threshold for the skewness and kurtosis features, respectively which are pre-calibrated.HO & H1 represents hypothesis test with LOS condition and NLOS condition resp.

OVERALL PERFORMANCE Static vs Mobile Links: The graph indicates that the

mobility increases spatial disturbance of NLOS paths.

Skewness vs Kurtosis: The kurtosis feature performs slightly better because skewness feature relies on extracting dominant paths.

Combining Skewness and Kurtosis: Combining the two feature and plotting the linear separator gives the marginal performance gain with LOS and NLOS detection rates of 94.36% and 95.98%

Fig : Overall LOS identification performance of the skewness and kurtosis features and their combination. (a) ROC curve of skewness and kurtosis. (b) ROCcurve of kurtosis.

Fig : Impact of propagation distances: Both features perform better for medium propagation ranges. (a) Skewness. (b) Kurtosis.

Impact of propagation distances

Fig : Impact of moving speed: Both features retain detection accuracy of above 82% for moving speeds of 0.5 m/s to 2.0 m/s. (a) Skewness. (b) Kurtosis.

Impact of moving speed

FUTURE SCOPE Target tracking

Wireless Energy Harvesting

LED – ID System

Mobile Radio Systems

PHY layer information to identify LOS conditions with wifi.

Explore skewness and kurtosis features for LOS identification.

Combination of skewness and kurtosis feature gives LOS and NLOS identification rates around 95%.

CONCLUSION

REFERENCES “WiFi-Based Indoor Line-of-Sight Identification” IEEE

Journel published in November 2015 – base paper J. Borras, P. Hatrack, and N. Mandayam, “Decision

theoretic framework for NLOS identification,” in Proc. IEEE Veh. Technol. Conf., 1998.

J. Lin, “Wireless power transfer for mobile applications, and health effects,” IEEE Antennas Propag. Mag., vol. 55, no. 2.

https://en.wikipedia.org/wiki/OSI_model

THANK YOU