A study on a correlation coefficient to associate an object with its user

Abstract

In this paper, we study a method of associating a person and an object utilized by him/her in real-time and an adhoc manner. The correlation coefficient is utilized to represent the proximity of the acceleration signal patterns from wearable sensors on different parts of the body and sensor augmented daily objects. The coefficients are evaluated in terms of the accuracy of the association and different varieties of calculating the correlation along 3D acceleration axes are studied. Also, the effects of the size of the time window and the sampling interval are examined.

A study on automatic recognition of object use exploiting motion correlation of wireless sensors

“…In this section, we compare the performance of several alternative motion correlation algorithms, including two described in the work by Fujinami et al [10,11]. We first compare the performance in terms of accuracy and response time, and we conclude this section with a comparison of the required computational effort for these algorithms.…”

“…Important advantages of RFID tags are that they are cheap and small and that no batteries are required. Unfortunately, spatial proximity information alone is often not enough to reliably detect the use of an object by a particular user, mainly because other nearby objects can also be detected [31], e.g., when the user moves in their proximity [10]. This problem is most prevalent when the communication range is large relative to the density of objects in the environment.…”

“…The work by Fujinami et al [10,11] explores using the correlation of raw accelerometer signals for object-user association. Much like our solution, Fujinami et al use the statistical correlation coefficient to establish an association between an object and a particular user.…”

“…One of the solutions proposed by Fujinami et al [10] to solve this problem is by taking the magnitude of the raw accelerometer signal. This way, only the intensity of the Figure 1a schematically shows an extreme example of the problem.…”

“…Several accelerometer-based algorithms are discussed in the initial paper by Fujinami et al [10], but we limit ourselves to the two most successful in this comparison: rawcompo and raw-max. Also, the work by Fujinami et al describes using multiple sensors spread over the body and choosing the best one to correlate with the artifact used.…”