The results obtained for the RFID prototype reveal that:

- For all cases analyzed, except for experiments 5 and 6, the RFID system stopped the machine automatically before the hand entered the dangerous area. This stop was initiated at an average distance of between 20 and 30 cm. In experiment 6 there were 10 occasions where the tag was detected but the hand was partially inside the dangerous region. By looking through the results from the stored data of this experiment: out of the 1200 distance detections recorded when the simulated hand passed through the dangerous zone, the RFID prototype output triggered 10 (0.83%) times when the hand was partially inside the dangerous region, and the minimum value was -132.3 mm, as shown in table 3. This value occurred only once, specifically in the entry point (3,7) of the matrix representation. As stated in Section 5, all the distance detections refer to the extreme point of the hand, which is at a distance of 180 mm from the tag (or tags). This means that the tag was detected at -132.3 mm + 180 mm = 47.7 mm from the plane defined by the antenna.

- The only case when no tag was detected is in experiment 5. In this case, out of the 1200 distance detections recorded, in 27 occasions no tag was detected, and in another 27 occasions the tag was detected when the simulated hand was partially inside the plane formed by the antenna (i.e. inside the dangerous region), which makes a total of 54 negative distances detected (4.5 %). In addition, out of 27 occasions where no tag was detected (and a penalty of -180 mm was used as stated in Section 6), 17 occurred in the entry point represented by the element (1,5) of the matrix representation. Moreover, out of the 27 occasions where the tag was detected but the simulated hand was partially inside the dangerous region, 9 occurred in that same entry point. This explains why the average value of the element (1,5) of the “Means” matrix in experiment 5 is negative.

- The results depend on the tag velocity. The matrices “Means” of Fig. 7 show that higher tag velocities result in tag detection at shorter distances. For example, the means of distance detections in experiment 1 for each entry point are within the interval of (270, 300) mm. These means are within the interval of (205, 245) mm in experiment 2, which was carried out at a linear velocity of 500 mm/s. In experiment 6 (velocity of 1600 mm/s), distance detection of the entry points are within the range of (130, 170) mm. This leads to the conclusion that velocity decreases distance detection.

- The distance detection depends on the orientation of the tag with respect to the antenna. This dependence is more severe with the presence of two tags. It was especially difficult to detect two tags at the same distance with the same orientation with respect to the antenna. Multitag detection capability would partially solve this problem and allow the detection of one or more tags inside the reading region.

- The accuracy obtained with this prototype in terms of repeatability of distance detections is poor. This is illustrated in the matrices “Means” and “Stds” in experiments 5 and 6 in Fig. 7, which show more disperse values compared to the other experiments conducted, with standard deviations within the interval (29, 60) mm. This, along with lack of multitag detection, is one of the main drawbacks of the system. Further studies must be undertaken to address these issues and to apply real time location techniques as addressed in recent works.

- By means of the experimental setup designed, it is possible to establish some tests to analyze the performance of security devices that carry out prevention functions. The benchmark tests presented here could serve to validate some regulations, such as EN 999 (1998). In addition, the matrix approach is flexible enough to define a set of different dangerous areas and thus to obtain statistics for each entry point defined. This flexibility permits to study the performance of safety devices in a set of different situations.