IEEE Consumer Electronics Magazine - March 2018 - 79

in their waveforms. However, their average, maximum, and minimum RSSI values are clearly different from those of other gestures. Gestures 4 and 5 (a hand moving toward and away from
the receiver) show a pattern of frequently changing RSSI values
at the end of the graph. Gesture 11 (a clap) exhibits a strong upand-down wave between the maximum value and minimum
value, and gesture 12 (a "come here" motion) shows a clear peak
and trough at the minimum value. In summary, hand gestures 1,
3, 4, 5, 11, and 12, can be recognized more easily compared to
other gestures by considering the changes of the Wi-Fi-RSSI
values. Therefore, these gestures can be used to design an application with an interactive interface.

EVALUATIONS OF THE MULTIPATH FADING
EFFECT AND LONG-DISTANCE ISSUE
Multipath fading is a phenomenon associated with wave
propagation, and it occurs when the receiver registers the
superposition of multiple copies of the transmitted signal,
with each copy traversing a different path. The results can be
either constructive (amplified) or destructive (attenuated)
interferences at the receiver. To investigate this phenomenon,
we performed 240 evaluations by considering three conditions and four gestures. For each gesture in a condition, 20
tests were run, and the conditions were as follows.
▼▼ Clean environment: no objects were around the sender and
receiver.
▼▼ Multipath fading by clutters: clutters were placed around
the sender and receiver.

Multipath fading by wall: the sender was placed at a distance of 10 cm from a wall.
The recognition methodology is described as follows:
having measured some RSSI values, the maximum likelihood estimation is used to evaluate the range of the RSSI
values. If the RSSI values (the mean, minimum, and maximum) are larger than the values under the no-gesture condition, then the user has performed a gesture. In this case, the
hidden Markov model is used to identify the gesture. More
specifically, in the hidden Markov model, a model for each
hand gesture was predefined by its characteristics, and the
various conditions were set as observations, with gestures
represented as hidden states. Thereafter, Viterbi algorithms
were used to match the sequence of the RSSI values with
predefined conditions, and the most probable hidden state
was the recognized gesture.
The outcomes are provided in Table 3, which shows that
the results of the environment with multipath fading are very
similar to those of the clean environment. Thus, the multipath
fading effect does not significantly affect the accuracy of our
recognition method. We also evaluated the performance at
four different distances using four gestures. For each gesture
in a distance, 20 tests were performed, and the results are
shown in Table 4. For the case of the longer distance, the signal strength level decreased. Thus, as shown in the table, the
longer distance reduces the prediction accuracy. Nevertheless, the system exhibits a good performance for distances
between 0.5 m and 2 m.

▼▼

Table 3. The accuracy of recognition of gestures under multipath fading environments.
Conditions

Prediction

Clean environment
Multipath fading by clutters
Multipath fading by a wall

No Gestures

Vertical Hand

Clap

Hand Moving Toward the Receiver

100%

85%

75%

25%

20%

25%

95%

80%

75%

70%

20%

15%

100%

80%

70%

75%

15%

20%

15%

Table 4. The accuracy of recognition of gestures at different distances.
Distance
0.5 m
1m
2m
3m

Prediction

No Gestures

Vertical Hand

Clap

Hand Moving Toward the Receiver

100%

85%

75%

25%

20%

25%

100%

75%

80%

75%

10%

20%

90%

75%

65%

70%

15%

25%

15%

80%

75%

50%

55%

10%

20%

25%

20%

MARCH 2018

IEEE Consumer Electronics Magazine

Table of Contents for the Digital Edition of IEEE Consumer Electronics Magazine - March 2018