Limits...
VisitSense: Sensing Place Visit Patterns from Ambient Radio on Smartphones for Targeted Mobile Ads in Shopping Malls.

Kim B, Kang S, Ha JY, Song J - Sensors (Basel) (2015)

Bottom Line: For accurate visit detection, we develop a change-based detection method to take into consideration the stability change of ambient radio and the mobility change of users.In addition, we proposed a causality-based visit prediction model to capture the causality in the sequential visit patterns for effective prediction.We have developed a VisitSense prototype system, and a visit-pattern-aware mobile advertising application that is based on it.

View Article: PubMed Central - PubMed

Affiliation: School of Computing, Korea Advanced Institute of Science and Technology (KAIST), 291 Daehak-ro, Yuseong, Daejeon 305-338, Korea. bjkim@nclab.kaist.ac.kr.

ABSTRACT
In this paper, we introduce a novel smartphone framework called VisitSense that automatically detects and predicts a smartphone user's place visits from ambient radio to enable behavioral targeting for mobile ads in large shopping malls. VisitSense enables mobile app developers to adopt visit-pattern-aware mobile advertising for shopping mall visitors in their apps. It also benefits mobile users by allowing them to receive highly relevant mobile ads that are aware of their place visit patterns in shopping malls. To achieve the goal, VisitSense employs accurate visit detection and prediction methods. For accurate visit detection, we develop a change-based detection method to take into consideration the stability change of ambient radio and the mobility change of users. It performs well in large shopping malls where ambient radio is quite noisy and causes existing algorithms to easily fail. In addition, we proposed a causality-based visit prediction model to capture the causality in the sequential visit patterns for effective prediction. We have developed a VisitSense prototype system, and a visit-pattern-aware mobile advertising application that is based on it. Furthermore, we deploy the system in the COEX Mall, one of the largest shopping malls in Korea, and conduct diverse experiments to show the effectiveness of VisitSense.

No MeSH data available.


An example of visit detection of VisitSense. Five adjacent places were visited. The black line represents real visits. High values (i.e., 1.5) of the black line indicates staying in a place, and low values (i.e., 0) indicates moving between places. The gray line represents detected visits. High values (i.e., 2.0) of the gray line indicates staying in a place, and low values (i.e., 0) indicates moving between places. The red line represents the Wi-Fi scan similarity measured by the Tanimoto coefficient, and its values rage from 0 to 1. The yellow line represents the SR of a SAL, and its values ranges from 0 to 1. The blue line represents the CR of a SAL, and its values range from 0 to 1. The green line represents the mobility measured as the RMS of accelerometer values in a time window. This example shows that visiting five adjacent places were well detected by using the proposed algorithm. The details of algorithm are explained in the Section 4.1.2.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4541934&req=5

sensors-15-17274-f008: An example of visit detection of VisitSense. Five adjacent places were visited. The black line represents real visits. High values (i.e., 1.5) of the black line indicates staying in a place, and low values (i.e., 0) indicates moving between places. The gray line represents detected visits. High values (i.e., 2.0) of the gray line indicates staying in a place, and low values (i.e., 0) indicates moving between places. The red line represents the Wi-Fi scan similarity measured by the Tanimoto coefficient, and its values rage from 0 to 1. The yellow line represents the SR of a SAL, and its values ranges from 0 to 1. The blue line represents the CR of a SAL, and its values range from 0 to 1. The green line represents the mobility measured as the RMS of accelerometer values in a time window. This example shows that visiting five adjacent places were well detected by using the proposed algorithm. The details of algorithm are explained in the Section 4.1.2.

Mentions: As mentioned earlier in this section, to detect entrance and departure, VisitSense monitors the ambient radio stability change and user mobility change. User mobility change is detected by accelerometer values. Using additional sensing modality, such as accelerometer effectively increases the detection accuracy by avoiding potential false positives due to erroneous changes of ambient radio. An example is provided in the following paragraphs (i.e., Departure Detection) by using Figure 8. VisitSense uses a three-axies accelerometer on a smartphone. VisitSense uses the sum of the Root-Mean-Square (RMS) value of accelerometer values in a time window (e.g., sampling rate: 5 Hz, window size: 1 min) for each axis. If the value is greater than the threshold, it means high mobility. If not, it is low mobility. Accordingly, we can detect changes in mobility. This heuristic algorithm was developed since the goal is to detect a user’s movements that may lead to the entrance or departure of a place. A simple algorithm that detects whether an accelerometer value is greater than a threshold at a certain time may be considered. However, we could not use such an algorithm, since it may detect false positives, such as in a situation where a user uses his/her smartphone at the same location.


VisitSense: Sensing Place Visit Patterns from Ambient Radio on Smartphones for Targeted Mobile Ads in Shopping Malls.

Kim B, Kang S, Ha JY, Song J - Sensors (Basel) (2015)

An example of visit detection of VisitSense. Five adjacent places were visited. The black line represents real visits. High values (i.e., 1.5) of the black line indicates staying in a place, and low values (i.e., 0) indicates moving between places. The gray line represents detected visits. High values (i.e., 2.0) of the gray line indicates staying in a place, and low values (i.e., 0) indicates moving between places. The red line represents the Wi-Fi scan similarity measured by the Tanimoto coefficient, and its values rage from 0 to 1. The yellow line represents the SR of a SAL, and its values ranges from 0 to 1. The blue line represents the CR of a SAL, and its values range from 0 to 1. The green line represents the mobility measured as the RMS of accelerometer values in a time window. This example shows that visiting five adjacent places were well detected by using the proposed algorithm. The details of algorithm are explained in the Section 4.1.2.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4541934&req=5

sensors-15-17274-f008: An example of visit detection of VisitSense. Five adjacent places were visited. The black line represents real visits. High values (i.e., 1.5) of the black line indicates staying in a place, and low values (i.e., 0) indicates moving between places. The gray line represents detected visits. High values (i.e., 2.0) of the gray line indicates staying in a place, and low values (i.e., 0) indicates moving between places. The red line represents the Wi-Fi scan similarity measured by the Tanimoto coefficient, and its values rage from 0 to 1. The yellow line represents the SR of a SAL, and its values ranges from 0 to 1. The blue line represents the CR of a SAL, and its values range from 0 to 1. The green line represents the mobility measured as the RMS of accelerometer values in a time window. This example shows that visiting five adjacent places were well detected by using the proposed algorithm. The details of algorithm are explained in the Section 4.1.2.
Mentions: As mentioned earlier in this section, to detect entrance and departure, VisitSense monitors the ambient radio stability change and user mobility change. User mobility change is detected by accelerometer values. Using additional sensing modality, such as accelerometer effectively increases the detection accuracy by avoiding potential false positives due to erroneous changes of ambient radio. An example is provided in the following paragraphs (i.e., Departure Detection) by using Figure 8. VisitSense uses a three-axies accelerometer on a smartphone. VisitSense uses the sum of the Root-Mean-Square (RMS) value of accelerometer values in a time window (e.g., sampling rate: 5 Hz, window size: 1 min) for each axis. If the value is greater than the threshold, it means high mobility. If not, it is low mobility. Accordingly, we can detect changes in mobility. This heuristic algorithm was developed since the goal is to detect a user’s movements that may lead to the entrance or departure of a place. A simple algorithm that detects whether an accelerometer value is greater than a threshold at a certain time may be considered. However, we could not use such an algorithm, since it may detect false positives, such as in a situation where a user uses his/her smartphone at the same location.

Bottom Line: For accurate visit detection, we develop a change-based detection method to take into consideration the stability change of ambient radio and the mobility change of users.In addition, we proposed a causality-based visit prediction model to capture the causality in the sequential visit patterns for effective prediction.We have developed a VisitSense prototype system, and a visit-pattern-aware mobile advertising application that is based on it.

View Article: PubMed Central - PubMed

Affiliation: School of Computing, Korea Advanced Institute of Science and Technology (KAIST), 291 Daehak-ro, Yuseong, Daejeon 305-338, Korea. bjkim@nclab.kaist.ac.kr.

ABSTRACT
In this paper, we introduce a novel smartphone framework called VisitSense that automatically detects and predicts a smartphone user's place visits from ambient radio to enable behavioral targeting for mobile ads in large shopping malls. VisitSense enables mobile app developers to adopt visit-pattern-aware mobile advertising for shopping mall visitors in their apps. It also benefits mobile users by allowing them to receive highly relevant mobile ads that are aware of their place visit patterns in shopping malls. To achieve the goal, VisitSense employs accurate visit detection and prediction methods. For accurate visit detection, we develop a change-based detection method to take into consideration the stability change of ambient radio and the mobility change of users. It performs well in large shopping malls where ambient radio is quite noisy and causes existing algorithms to easily fail. In addition, we proposed a causality-based visit prediction model to capture the causality in the sequential visit patterns for effective prediction. We have developed a VisitSense prototype system, and a visit-pattern-aware mobile advertising application that is based on it. Furthermore, we deploy the system in the COEX Mall, one of the largest shopping malls in Korea, and conduct diverse experiments to show the effectiveness of VisitSense.

No MeSH data available.