Analyzing Shopper's Behavior through WiFi Signals

Abstract

Substantial progress in WiFi-based indoor localization has proven that pervasiveness of WiFi can be exploited beyond its traditional use of internet access to enable a variety of sensing applications. Understanding shopper's behavior through physical analytics can provide crucial insights to the business owner in terms of e↵ectiveness of promotions, arrangement of products and e ciency of services. However, analyzing shopper's behavior and browsing patterns is challenging. Since video surveillance can not used due to high cost and privacy concerns, it is necessary to design novel techniques that can provide accurate and e cient view of shopper's behavior. In this work, we propose WiFi-based sensing of shopper's behavior in a retail store. Specifically, we show that various states of a shopper such as standing near the entrance to view a promotion or walking quickly to proceed towards the intended item can be accurately classified by profiling Channel State Information (CSI) of WiFi. We recognize a few representative states of shopper's behavior at the entrance and inside the store, and show how CSI-based profile can be used to detect that a shopper is in one of the states with very high accuracy (⇡ 90%). We discuss the potential and limitations of CSI-based sensing of shopper's behavior and physical analytics in general.

Robust Indoor Human Activity Recognition Using Wireless Signals

“…CSIs in different locations have been explored by many studies to locate people in a room and have had positive results [10,14,18]. However, we want the features in our method to be insensitive to the locations.…”

“…Studies [10,11] have proved that the existence and movement of humans will affect the channel state information (CSI) of wireless signals, and CSI has an advantage over light, infrared, or thermal energy when attempting to infer people’s movements. CSI holds potential for the convergence of accurate and pervasive indoor localization and has attracted numerous recent research efforts [10,11,12,13,14]. Studies [10,11,12] have shown that different actions have different CSI change patterns.…”

“…E-eyes [13] can distinguish a set of in-place human activities with only a single Wi-Fi access point. Paper [14] can accurately classify five states of the shopper during a typical in-store visit.…”

Towards Location Independent Gesture Recognition with Commodity WiFi Devices

“…CARM [50] body movements 0.15∼300 WiSee [23] hand gestures 8∼134 WiFinger [27] hand gestures 0.2∼5 WiFinger [26] ASL gestures 1∼60 WiWho [47] gaits 0.3∼2 WiFi-ID [51] walking behavior 20 ∼80 UniBreathe [52] respiration 0.1∼0.5 WiHear [53] mouth 2∼5 Zeng [54] shopping behavior 0.3∼2…”

BeSense: Leveraging WiFi Channel Data and Computational Intelligence for Behavior Analysis

“…set(j));19 for x = set(j) : length(nor2) do20 if nor2(x) <= e then21 end-point=x; max(w) − min(w) < threshold then29 (a) Original CSI waveform of 17 consecutive typing gestures; (b) nor1: variance of original waveform ; (c) nor2: data smoothing on nor1; (d) Comparison of nor1 and nor2; (e) Start-point determination for 3rd gesture in original waveform. Since the result tends to be stable after the 7th segment, set the 7th segment point as start-point.…”