Dynamic spectrum sharing is regarded as an emerging technique for next generation wireless systems, where unlicensed secondary users opportunistically access RF channels, which are not licensed to them but to primary users. Furthermore, dynamic spectrum sharing requires that the unlicensed secondary users should not create any harmful interference to licensed primary users while accessing RF channels opportunistically. Thus, unlicensed secondary users need perfect information about whether any licensed primary users are active in a channel of interest.
There are several spectrum sensing algorithms to identify whether the given channel is idle (not used by primary users) or busy (used by primary users). Most commonly used algorithm is energy detection (as it does not requires knowledge of primary user signal), where spectrum sensor calculates the energy of the received signal in a given channel and compares it with a pre-specified threshold value. If the energy of the received signal is greater than the threshold value, spectrum sensor decides that the given channel is busy. Otherwise it decides that the channel is idle.
However, it is noted that when only energy based detection is used, it leads to high false detection rate and/or misdetection. Thus, in order to reduce the false detection and misdetection rates, spectrum sensor needs to consider both energy peak and width of the signal spectra. To illustrate this scenario, we consider following figure. Figure below was the outcome of spectrum sensor when two of the 5 GHz Wi-Fi channels (5.26GHz and 5.66 GHz channels) were active for a given time and location. In the figure, we can see that when Threshold 1 (-75 dBm, red line) was used, the channel 5.66 GHz channel was not detected. Furthermore, when Threshold 2 (-85 dBm, green line) was used, both channels 5.26GHz and 5.66 GHz channels were detected. In addition to these two channels, two other channels (at 5.2 GHz and 5.5 GHz) were detected which were not supposed to be detected as active. When spectrum sensor used only energy based detection, 2 extra false channels were detected. However, when both width of the signal spectra (~ 60% of bandwidth of a Wi-Fi channel) and energy level were considered while detecting the primary user signal, only truly active channels, 5.26GHz and 5.66 GHz, were detected as active, while two other channels, 5.2 GHz and 5.5 GHz, were not detected as active. Thus, joint energy and width of signal spectra detection approach may reduce the spectrum uncertainty to avoid any harmful interference to primary users, while secondary users get accurate channel occupancy information for opportunistic communications.
